The results suggest that Oregon transportation agencies might improve driver-yielding behavior .... Trademarks or manufacturers' names appear herein only because they are ...... Periodic enforcement may be needed ...... Oregon law the expected correct action for both display indications is the same for right turns. A.
IMPROVED SAFETY AND EFFICIENCY OF PROTECTED/PERMITTED RIGHTTURNS IN OREGON Final Report PROJECT SPR 789
IMPROVED SAFETY AND EFFICIENCY OF PROTECTED/PERMITTED RIGHT-TURNS IN OREGON Final Report PROJECT SPR 789 by David Hurwitz, Ph.D., Associate Professor Hisham Jashami, Graduate Research Assistant Kamilah Buker, Graduate Research Assistant Oregon State University School of Civil and Construction Engineering 101 Kearney Hall Corvallis, OR 97331 Christopher Monsere, Ph.D., P.E., Associate Professor Sirisha Kothuri, Ph.D., Research Associate Andrew Kading, Graduate Research Assistant Department of Civil and Environmental Engineering Portland State University PO Box 751 Portland, OR 97207 for Oregon Department of Transportation Research Section 555 13th Street NE Salem OR 97301
May 2018
Technical Report Documentation Page
1. Report No.
2. Government Accession No.
3. Recipient’s Catalog No.
FHWA-OR-RD-18-14 4. Title and Subtitle Improved Safety and Efficiency of Protected/Permitted Right-Turns in Oregon
7. Author(s) David Hurwitz, Chris Monsere, Sirisha Kothuri, Hisham Jashami, Kamilah Buker, Andrew Kading
9. Performing Organization Name and Address Oregon State University Corvallis, Oregon 97331
8. Performing Organization Report No. 10. Work Unit No. (TRAIS)
Portland State University Portland OR 97207
11. Contract or Grant No.
12. Sponsoring Agency Name and Address Oregon Dept. of Transportation Research Section and 555 13th Street NE, Suite 1 Salem, OR 97301
5. Report Date May 2018 6. Performing Organization Code
13. Type of Report and Period Covered
Federal Highway Admin. 400 Seventh Street, SW Washington, DC 20590-0003
Final Report 14. Sponsoring Agency Code
15. Supplementary Notes 16. Abstract This research aimed to develop an understanding of the safety and operational implications of using the flashing yellow arrow (FYA) in permitted and protected/permitted right turn (PPRT) operations to maximize safety and efficiency. This report includes a review of more than 50 scientific and technical articles on the selection of right turn phasing alternatives and a review of Oregon crash data at intersections with exclusive right-turning lanes from 2011-2013. The study had three phases: 1) a web-based survey, 2) a microsimulation model, and 3) and a driving simulator study. The survey measured Oregon driver’s comprehension of right-turn signal display. Analysis revealed a general misunderstanding of the required driver response for the steady red arrow signal indication, but comprehension of the FYA for right turns was high. The microsimulation model of several PPRT phasing alternatives indicated that the pedestrian volumes had the greatest effects on delays. The driving simulator experiment indicated that driver responses were relatively consistent with those observed in the web-based survey. The results suggest that Oregon transportation agencies might improve driver-yielding behavior and pedestrian safety at signalized intersections with high volumes of permissive right turns from exclusive right-turn lanes by using the FYA display in lieu of a steady circular green display.
17. Key Words
18.
Distribution Statement
Protected/Permitted Right Turns, Safety, Human Factors, Traffic Operations, PPRT, Flashing Yellow Arrow (FYA) 19. Security Classification (of this report) Unclassified Technical Report Form DOT F 1700.7 (8-72)
20. Security Classification (of this page) Unclassified Reproduction of completed page authorized
i
21. No. of Pages 22. Price
Printed on recycled paper
ii
SI* (MODERN METRIC) CONVERSION FACTORS APPROXIMATE CONVERSIONS TO SI UNITS Symbol
When You Know
Multiply By
To Find
APPROXIMATE CONVERSIONS FROM SI UNITS
Symbol
Symbol
When You Know
LENGTH in ft yd mi
inches feet yards miles
25.4 0.305 0.914 1.61 AREA
in2
square inches
645.2
ft2 yd2 ac
square feet square yards acres
0.093 0.836 0.405
mi2
square miles
2.59
To Find
Symbol
LENGTH millimeters meters meters kilometers
mm m m km
mm m m km
millimeters meters meters kilometers
millimeters squared meters squared meters squared hectares kilometers squared
mm2
mm2
m2 m2 ha
m2 m2 ha
km2
km2
millimeters squared meters squared meters squared hectares kilometers squared
VOLUME fl oz fluid ounces 29.57 milliliters ml gal gallons 3.785 liters L ft3 cubic feet 0.028 meters cubed m3 yd3 cubic yards 0.765 meters cubed m3 NOTE: Volumes greater than 1000 L shall be shown in m3. MASS oz ounces 28.35 grams g lb pounds 0.454 kilograms kg short tons (2000 T 0.907 megagrams Mg lb)
ml L m3 m3
milliliters liters meters cubed meters cubed
0.039 3.28 1.09 0.621 AREA
inches feet yards miles
in ft yd mi
0.0016
square inches
in2
10.764 1.196 2.47
square feet square yards acres
ft2 yd2 ac
0.386
square miles
mi2
VOLUME 0.034 fluid ounces 0.264 gallons 35.315 cubic feet 1.308 cubic yards
g kg
grams kilograms
MASS 0.035 2.205
Mg
megagrams
1.102
TEMPERATURE (exact)
fl oz gal ft3 yd3
ounces pounds
oz lb
short tons (2000 lb)
T
TEMPERATURE (exact)
(FCelsius °C 32)/1.8 *SI is the symbol for the International System of Measurement °F
Multiply By
Fahrenheit
°C
iii
Celsius
1.8C+3 2
Fahrenheit
°F
iv
ACKNOWLEDGEMENTS The authors thank the Oregon Department of Transportation (ODOT) and the Federal Highway Administration (FHWA) for funding this research. The Technical Advisory Committee has provided valuable input throughout the project (Craig Black-ODOT, Scott Cramer-ODOT, Julie Kentosh-ODOT, Katie Johnson-ODOT, and Bikram Raghubansh-Clackamas County). Additional students at Oregon State University helped reduce data and code the simulator environment (Kayla Fleskes, MS, Ellie Simpson, MS, Hameed Aswad Mohammed, PhD, Hagai Tapiro, Post-Doctoral researcher, and Logan Scott-Deeter, undergraduate). The project was managed expertly by Mark Joerger, ODOT Research Coordinator.
DISCLAIMER This document is disseminated under the sponsorship of the Oregon Department of Transportation and the United States Department of Transportation in the interest of information exchange. The State of Oregon and the United States Government assume no liability of its contents or use thereof. The contents of this report reflect the views of the author(s) who are solely responsible for the facts and accuracy of the data presented herein. The contents do not necessarily reflect the official policies of the Oregon Department of Transportation or the United States Department of Transportation. The State of Oregon and the United States Government do not endorse products of manufacturers. Trademarks or manufacturers’ names appear herein only because they are considered essential to the object of this document. This report does not constitute a standard, specification, or regulation.
v
vi
TABLE OF CONTENTS 1
INTRODUCTION................................................................................................................. 1
2
LITERATURE REVIEW .................................................................................................... 3 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 2.10
3
ANALYSIS OF CRASH DATA ........................................................................................ 15 3.1
4
TYPES OF TURNING MOVEMENTS .................................................................................... 3 GEOMETRY ...................................................................................................................... 3 PHASING .......................................................................................................................... 5 DISPLAY........................................................................................................................... 8 USE OF DETECTORS IN RIGHT-TURN OPERATIONS ........................................................... 9 EFFECTS OF RESTRICTING RIGHT TURNS ON RED .......................................................... 10 PEDESTRIANS ................................................................................................................. 10 BICYCLISTS .................................................................................................................... 12 SAFETY .......................................................................................................................... 12 SUMMARY ...................................................................................................................... 13 SUMMARY ...................................................................................................................... 19
SURVEY .............................................................................................................................. 21 4.1 DESIGN AND REFINEMENT ............................................................................................. 21 4.2 INSTRUMENT .................................................................................................................. 21 4.3 ADMINISTRATION .......................................................................................................... 22 4.4 RESPONSE RATE............................................................................................................. 24 4.5 DEMOGRAPHIC SUMMARY ............................................................................................. 25 4.6 RESULTS AND ANALYSIS................................................................................................ 27 4.6.1 Coding ....................................................................................................................... 27 4.6.2 Open-ended Comprehension Questions .................................................................... 30 4.6.2.1 Steady Circular Green Ball .............................................................................. 30 4.6.2.2 Steady Green Arrow .......................................................................................... 31 4.6.2.3 Steady Circular Red .......................................................................................... 32 4.6.2.4 Steady Red Arrow ............................................................................................. 33 4.6.2.5 Flashing Yellow Arrow ..................................................................................... 34 4.6.3 Multiple-Choice Comprehension Questions ............................................................. 35 4.6.3.1 Steady Red Arrow Comprehension ................................................................... 35 4.6.3.2 Red Arrow and Circular Red Comparison ....................................................... 36 4.6.3.3 Flashing Yellow Arrow Comprehension ........................................................... 37 4.6.3.4 Flashing Yellow Arrow, Circular Green Comparison...................................... 38 4.7 SUMMARY ...................................................................................................................... 39
5
MICROSIMULATION ...................................................................................................... 41 5.1 SIMULATION MODEL BACKGROUND .............................................................................. 41 5.2 TYPES OF SIMULATION MODELS .................................................................................... 41 5.3 MODEL DEVELOPMENT .................................................................................................. 42 5.3.1 Site Selection ............................................................................................................ 43 5.3.2 Geometry................................................................................................................... 43 5.3.3 Volumes .................................................................................................................... 45
vii
5.3.4 Vehicle Composition ................................................................................................ 45 5.3.5 Signal Timing............................................................................................................ 45 5.4 RESULTS ........................................................................................................................ 46 5.4.1 Base Model ............................................................................................................... 46 5.4.2 Sensitivity Analysis .................................................................................................. 47 5.5 SUMMARY ...................................................................................................................... 48 6
DRIVING SIMULATOR EXPERIMENT ....................................................................... 49 6.1 EXPERIMENTAL EQUIPMENT........................................................................................... 49 6.1.1 Driving Simulator ..................................................................................................... 49 6.1.1.1 Simulator Data .................................................................................................. 51 6.1.1.2 Simulator Sickness ............................................................................................ 52 6.1.2 Eye Tracker ............................................................................................................... 52 6.2 EXPERIMENTAL DESIGN ................................................................................................. 53 6.2.1 Road and Intersection Geometry .............................................................................. 54 6.2.2 Experimental Variables ............................................................................................. 55 6.2.2.1 Independent Variables ...................................................................................... 55 6.2.2.2 Dependent Variables ......................................................................................... 56 6.2.3 Factorial Design ........................................................................................................ 56 6.2.4 Counterbalancing and Presentation of Driving Scenarios ........................................ 57 6.3 DRIVING SIMULATOR EXPERIMENTAL PROTOCOL ......................................................... 59 6.3.1 Recruitment ............................................................................................................... 59 6.3.2 Informed Consent and Compensation....................................................................... 60 6.3.3 Prescreening Survey.................................................................................................. 60 6.3.4 Calibration Drive ...................................................................................................... 61 6.3.5 Eye-Tracking Calibration.......................................................................................... 62 6.3.6 Experimental Drive ................................................................................................... 63 6.3.7 Postdrive Survey ....................................................................................................... 63 6.4 DATA REDUCTION ......................................................................................................... 63 6.4.1 Eye-Tracking Data .................................................................................................... 63 6.4.2 Simulator Data .......................................................................................................... 65 6.5 RESULTS AND ANALYSIS .............................................................................................. 65 6.5.1 Participants ................................................................................................................ 66 6.5.1.1 Summary Statistics ............................................................................................ 66 6.5.1.2 Demographics ................................................................................................... 66 6.5.2 Postdrive Survey Results .......................................................................................... 67 6.5.3 Driver Decision Making ........................................................................................... 69 6.5.3.1 Steady Circular Green ...................................................................................... 72 6.5.3.2 Steady Green Arrow .......................................................................................... 72 6.5.3.3 Steady Circular Red .......................................................................................... 73 6.5.3.4 Steady Red Arrow ............................................................................................. 73 6.5.3.5 Flashing Yellow Arrow with Pedestrian Walk Interval (FYAW) ...................... 73 6.5.3.6 Flashing Yellow Arrow with Pedestrian Clearance Interval (FYAC) .............. 74 6.5.4 Visual Attention ........................................................................................................ 74 6.5.4.1 Average Total Fixation Duration (ATFD) ........................................................ 74 6.5.4.2 Descriptive Data and Statistical Analysis......................................................... 74 6.5.4.3 Circular Red and Red Arrow Indications ......................................................... 75 viii
Circular Green and Flashing Yellow Arrow Indications ................................. 76 Flashing yellow arrow with walk interval and with clearance interval ........... 78 Presence of Pedestrians .................................................................................... 80 6.6 SUMMARY ...................................................................................................................... 81 6.6.1 Driver Decision Making ........................................................................................... 81 6.6.2 Driver Visual Attention............................................................................................. 81 6.5.4.4 6.5.4.5 6.5.4.6
7
CONCLUSIONS ................................................................................................................. 83 7.1 SURVEY ......................................................................................................................... 83 7.2 MICROSIMULATION ........................................................................................................ 84 7.3 DRIVING SIMULATOR ..................................................................................................... 85 7.3.1 Driver Decision Making ........................................................................................... 85 7.3.2 Driver Visual Attention............................................................................................. 85 7.4 SYNTHESIS AND DISCUSSION OF RESULTS ..................................................................... 86 7.4.1 Indications That Require the Driver to Yield ........................................................... 86 7.4.2 Indications That Require Drivers to Stop ................................................................. 87 7.4.3 Indications That Communicate to the Driver That the Movement Is Exclusive ...... 87 7.5 LIMITATIONS OF THE RESEARCH .................................................................................... 88 7.6 RECOMMENDATIONS FOR PRACTICE .............................................................................. 88
8
REFERENCES .................................................................................................................... 91
APPENDIX A APPENDIX B APPENDIX C APPENDIX D
NHTSA DATA SURVEY INSTRUMENT SAMPLE OPEN ENDED RESPONSES GRID LAYOUTS AND FIGURES OF SURVEY-SIMULATOR RESPONSES
ix
List of Tables Table 2.1: Potential benefits and liabilities of right-turn lanes (FHWA 2015) ..............................................................5 Table 2.2: Typical right-turn overlap settings (FHWA 2015) .......................................................................................7 Table 3.1: Summary of ODOT crash data (2011–2013).............................................................................................. 15 Table 3.2 Severity of ODOT right-turn crashes at signalized intersections (2011–2013) ........................................... 15 Table 3.3: Comparison of proportions of crashes at intersections by collision type.................................................... 17 Table 3.4: Comparison of proportions of crash types at intersections by cause of crash............................................. 18 Table 4.1: Sampling scheme for driver survey ............................................................................................................ 23 Table 4.2 Demographic comparison between survey and Census respondents ........................................................... 26 Table 4.3: Characteristics of survey respondents ........................................................................................................ 27 Table 4.4: Error Coding of Narrative........................................................................................................................... 28 Table 4.5: Estimated values of Cohen’s kappa coefficient .......................................................................................... 30 Table 4.6: Responses to open-ended question on steady green circular ball signal indication .................................... 31 Table 4.7: Responses to open-ended question on steady green arrow signal indication.............................................. 32 Table 4.8: Responses to open-ended question on steady circular red signal indication .............................................. 33 Table 4.9: Responses to open-ended question on steady red arrow signal indication ................................................. 34 Table 4.10: Responses to open-ended question on flashing yellow arrow signal indication ....................................... 35 Table 4.11: Responses to multiple-choice question on steady red arrow signal indication ......................................... 36 Table 4.12: Confidence levels for responses to multiple-choice question on steady red arrow signal indication ....... 36 Table 4.13: Comprehension of red arrow vs. circular red signal indications............................................................... 37 Table 4.14: Responses to multiple-choice question on flashing yellow arrow indication ........................................... 38 Table 4.15: Confidence level for responses to multiple-choice question on flashing yellow arrow indication ........... 38 Table 4.16: Comprehension of flashing yellow arrow vs. circular green signal indications ....................................... 39 Table 5.1 Average green times from simulation .......................................................................................................... 47 Table 5.2 Percent change in delay, by movement and phase ....................................................................................... 48 Table 6.1: Experimental variables and levels .............................................................................................................. 57 Table 6.2: Grids and right-turning intersection layout ................................................................................................ 58 Table 6.3: Summary of participant population ............................................................................................................ 66 Table 6.4: Participant demographics ........................................................................................................................... 67 Table 6.5: Postdrive survey question responses .......................................................................................................... 69 Table 6.6: Error coding of narrative for the simulator experiments ............................................................................ 70 Table 6.7: Steady circular green circular responses..................................................................................................... 72 Table 6.8: Steady green arrow responses..................................................................................................................... 72 Table 6.9: Steady circular red responses ..................................................................................................................... 73 Table 6.10: Steady red arrow responses ...................................................................................................................... 73 Table 6.11: Responses to flashing yellow arrow during pedestrian walk interval ....................................................... 73 Table 6.12: Responses to flashing yellow arrow during pedestrian clearance interval................................................ 74 Table 6.13: Descriptive statistics for TFD by signal type, bay length, and AOI ......................................................... 76 Table 6.14: Tukey’s HSD multiple comparisons by signal type, bay length, and AOI ............................................... 76 Table 6.15: Descriptive statistics for TFD by signal type, bay length, and AOI ......................................................... 77 Table 6.16: Tukey’s HSD multiple comparisons by signal type, bay length, and AOI ............................................... 78 Table 6.17: Descriptive statistics for TFD by signal type, bay length and AOI .......................................................... 79 Table 6.18: Tukey’s HSD multiple comparisons by signal type, bay length, and AOI ............................................... 80 Table 6.19: Two-sample t-test of ATFDs comparing AOIs by presence of conflicting pedestrian ............................. 80
x
List of Figures Figure 2.1: Island geometry (ODOT) ............................................................................................................................4 Figure 2.2: General phasing with permitted phasing for right turns (USDOT 2015) ....................................................6 Figure 2.3: Phasing scheme for right-turn overlaps (USDOT 2015) .............................................................................7 Figure 2.4 Oregon Signal Head Types for Right-Turns.................................................................................................8 Figure 2.5: Example scheme for PPRT phasing (ODOT 2015).....................................................................................9 Figure 2.6: Entrance (left) and exit (right) treatments (Dixon et al. 2000) .................................................................. 13 Figure 4.1: Recruitment postcard for the survey ......................................................................................................... 24 Figure 4.2: Geographic distribution of respondents .................................................................................................... 25 Figure 4.3: Image used for open-ended question on steady green circular ball signal indication (with RTO sign) .... 31 Figure 4.4: Image used for open-ended question on steady green arrow signal indication (with RTO sign) .............. 32 Figure 4.5: Image used for open-ended question on steady circular red signal indication (with RTO sign) ............... 33 Figure 4.6: Image used for open-ended question on steady red arrow signal indication (with RTO sign) .................. 34 Figure 4.7: Image used for open-ended question on flashing yellow arrow signal indication (with RTO sign) ......... 35 Figure 4.8: Image used for question on comprehension of red arrow vs. circular red signal indications .................... 37 Figure 4.9: Image used for question on comprehension of flashing yellow arrow vs. circular green signal indications ........................................................................................................................................................................... 39 Figure 5.1: HILS data flow (Stevanovic et al. 2009) ................................................................................................... 42 Figure 5.2: Data flow in SITL model (Stevanovic et al. 2009) ................................................................................... 42 Figure 5.3: Site location for simulation ....................................................................................................................... 43 Figure 5.4: Intersection geometry ................................................................................................................................ 44 Figure 5.5: Volumes used in the simulation model ..................................................................................................... 44 Figure 5.6: Phase rotation diagram for NW Cornell Rd. and NW Evergreen Pkwy.................................................... 45 Figure 5.7: Average delays by movement ................................................................................................................... 46 Figure 6.1: Operator workstation for driving simulator. Left: Designing an experiment in the Internet Scene Assembler with Javascript. Right: Researcher evaluating a newly designed environment ................................ 49 Figure 6.2: Simulated environment in OSU driving simulator, from the participant’s perspective inside (left) and from outside (right) the vehicle .......................................................................................................................... 50 Figure 6.3: Operator workstation for full-scale driving simulator. Monitors are shown displaying SimObserver (left), and the simulated environment (center) or vehicle dashboard (right) as seen in the vehicle ............................. 51 Figure 6.4: Screenshots of the six views from SimObserver. Clockwise from top left: simulated scene as projected on the screen; driver’s upper body and hands on steering wheel; acceleration and brake pedals in vehicle; entire simulator viewed from outside the vehicle; steering wheel and dashboard; and driver’s face................. 52 Figure 6.5: OSU researcher demonstrating the Mobile Eye XG Glasses (left) and Mobile Recording Unit (right).... 53 Figure 6.6: Example of signal and pedestrian head configurations in the simulator environment .............................. 53 Figure 6.7: Screenshot of a sample environment coded in the simulator .................................................................... 54 Figure 6.8: Screenshot of a cross-section of the right-turn roadway environment coded in the simulator .................. 54 Figure 6.9: Example scenario of a participant approaching the right turn on a red arrow ........................................... 55 Figure 6.10: Example scenario of right turns with 50- and 100-ft turning bay lengths ............................................... 56 Figure 6.11: Example scenario of right turns with and without the presence of pedestrians ....................................... 56 Figure 6.12: Example layout of Grid 1 with four right-turning (RT) scenarios – Path Start-Right-Right-RightThrough-Right-Right-Left-Left-Right-Finish .................................................................................................... 59 Figure 6.13: Layout of calibration drive ...................................................................................................................... 61 Figure 6.14: Screenshots of calibration drive in the simulation. Left: Approach to a curve near the beginning of the drive. Right: Approach to a signalized intersection, where a yellow billboard indicates a right turn ................ 62 Figure 6.15: Eye-tracking calibration image ............................................................................................................... 62 Figure 6.16: Example of a participant fixation pattern on the signal head (two AOIs) ............................................... 64 Figure 6.17: Example of a participant fixation pattern on pedestrian (three AOIs) ..................................................... 64 Figure 6.18: Screenshot of Data Distillery software interface (identifiable participant information was removed) ... 65 Figure 6.19: Participant comprehensions of the flashing yellow arrow (FYA) and red arrow (RA) indications ........ 68 Figure 6.20: Screenshot of Simobserver platform ....................................................................................................... 71 Figure 6.21: Boxplots of TFD disaggregated by signal type ....................................................................................... 75 Figure 6.22: Boxplots of TFD disaggregated by signal type ....................................................................................... 77 Figure 6.23: Boxplots of TFD disaggregated by signal type ....................................................................................... 79
xi
Figure 7.1: Signal Heads for Right-Turn Operations ................................................................................................... 90
xii
1 INTRODUCTION Accommodating motor vehicles that are turning (left or right) at signalized intersections requires a careful understanding of the safety and efficiency of design and operational variables. Turning vehicles are the primary collision risk for non-motorized road users. When turning movements need to be controlled directly, proper driver response to traffic control is critical. A substantial and compelling body of research has established that the flashing yellow arrow indication for permitted left turns improves driver comprehension and behavioral responses to the permissive condition (yield to other vehicles and persons in conflict with movement) and highlights the potential risks for pedestrians. Because right turns are generally permitted during the pedestrian walk and clearance indications, it is common for right-turning drivers to make yielding errors. Although traffic engineers have a good understanding of driver comprehension and response to the circular green ball or solid green arrow indications for right-turning movements, there is limited research related to use of the flashing yellow arrow indication for right-turn movements or driver comprehension of the solid red right arrow in Oregon. The objective of this research was to develop an understanding of the safety and operational implications of using the flashing yellow arrow indication to indicate a permitted right turn. Following a comprehensive review of the literature and Oregon crash data, which identified relevant factors such as roadway geometry, traffic, and pedestrian volumes, the research focused on intersections with exclusive right-turning lanes. The research used three key efforts to accomplish these objectives: •
A web-based survey to understand the comprehension of Oregon drivers of right-turn signal display alternatives (see Chapter 4);
•
A microsimulation to explore the operational performance of several protected/permitted right-turn (PPRT) phasing alternatives under varying volumes (right-turn vehicle volumes, conflicting pedestrian movements, and conflicting left-turn vehicular movements) (see Chapter 5); and
•
A driving simulator experiment to examine motorist behavior in response to right-turn signal displays with two levels of both pedestrian activity and length of turning bays (Chapter 6).
This Final Report summarizes the research and is organized into eight chapters. Chapter 2 presents a brief literature review. Chapter 3 presents the analysis of crash data at signalized intersections in Oregon. Chapter 4 describes the survey administration and analysis of the results. Chapter 5 presents the software-in-the-loop microsimulation models. Chapter 6 describes the design of the driving simulator experiment and results of the analysis. Chapter 7 summarizes the findings of the major research tasks, synthesizes the results, and presents recommendations for the use of PPRT phasing to maximize the safety of non-motorized road users and the overall efficiency of ODOT’s signalized intersections. Cited references are summarized in Chapter 8.
1
2
2 LITERATURE REVIEW This chapter reviews the literature, including design manuals, guidance documents, and published articles, related to right-turn operations at signalized intersections. The chapter is organized by topical area and concludes with a summary. No research was found that evaluated flashing yellow arrow use in the context of protected-permissive right-turn (PPRT) operations.
2.1
TYPES OF TURNING MOVEMENTS
Right turns at a signalized intersections can be categorized as either: (1) right turns that have the right of way (ROW), or (2) right turns that must yield, to be consistent with the rules of the road (USDOT 2015). A protected right turn falls into the first category of right-turn movements: the ROW is provided, and no conflicting vehicles (or pedestrians) are allowed (USDOT 2015). A permissive right turn falls into the second category: drivers are only allowed to proceed through the intersection if there is an acceptable gap in the conflicting flow of vehicles, including bicycles, or pedestrians (USDOT 2015). A protected plus permitted turn is a combination that either begins with a protected movement and transitions into a permitted movement or vice versa. (USDOT 2015).
2.2
GEOMETRY
An protected right turn requires a dedicated right-turn lane that is used only by vehicles making right turns (USDOT 2015). Intersection approach designs must accommodate turning movements for all appropriate design vehicles. Turning radius is a key element in intersection design (ODOT 2012). Before designing the intersection, the appropriate design vehicle must be identified. The intersection radius should be selected to minimize the overall size of the intersection and the pedestrian crossing distance (ODOT 2012). For example, if most of the turning traffic comprises passenger cars, then it is not cost-effective or pedestrian-friendly to design the turning radius for large trucks (AASHTO 2011). Through-traffic and right-turning traffic have differential speeds and can potentially cause safety issues in a shared lane configuration. Speed differentials in a shared lane can result in increased delay for through vehicles and increased likelihood of rear-end crashes (FHWA 2004). To mitigate this problem, use of an exclusive right-turn lane may be appropriate (ODOT 2012). Exclusive right-turn lanes improve safety and have the potential to improve the overall operation and efficiency of the intersection (ODOT 2012). Right-turn lanes are generally designed to be 12-ft wide with a shoulder of 3–4 ft (ODOT 2012). When designing an exclusive right-turn lane in an urban environment, through-bike movement needs to be incorporated (ODOT 2012). Conflicts between right-turning traffic and bicyclists can be minimized by positioning the bike lane to the left of the right-turn lane, to align the bicyclist better with the through movement as the bicyclist travels downstream (ODOT 2012). The design of a bike lane between a through lane and a right-turn lane contains a conflict point where the 3
cyclist’s path must be crossed by a right-turning vehicle further back from the intersection (ODOT 2012). To alert users to conflicting paths in this area, the bike lane is usually marked with short skip striping (ODOT 2012). As cited in the Oregon Department of Transportation (ODOT) Traffic Signal Policy and Guidelines, “The Manual on Uniform Traffic Control Devices (MUTCD) and the Federal Highway Administration (FHWA) have allowed this area to be colored green as an experimental condition to draw more attention to the conflict area” (ODOT 2012). The standard width of a bike lane located between the through and right-turn lanes is 5 ft (ODOT 2012). Turning roadways and channelization are important aspects of intersection design (AASHTO 2011). For intersections, turning roadway widths are based on the volume and classification of turning vehicles that need to be accommodated. Most turning roadways are designed for rightturning traffic. Three types of right-turning roadways are possible at intersections: a corner triangular island design, a minimum edge-of-travel-way design, and a free-flow design utilizing a simple radius or compound radii (AASHTO 2011). The American Association of State Highway and Transportation Officials (AASHTO) defines “channelization” as the separation or regulation of “conflicting traffic movements into definite paths or travel by traffic islands or pavement markings to facilitate the orderly movements of both vehicles and pedestrians” (AASHTO 2011). Use of channelization increases the capacity and operational efficiency, reduces the crash frequency, and provides positive guidance to motorists (AASHTO 2011).
Figure 2.1: Island geometry (ODOT) AASHTO defines an island as the area used for the control of movements performed by vehicles between traffic lanes (AASHTO 2011). Channelization islands are used to protect pedestrians, store traffic control devices, separate conflicts, control the angle of conflict, eliminate excessive pavement areas, regulate traffic, favor predominant turning movements, and protect or localize turning or crossing vehicles (AASHTO 2011). Islands can improve operations and safety by separating right-turning and through movements (FHWA 2004). Raised or painted islands can provide large turning radii to accommodate larger vehicles, thereby enabling faster turning speeds (FHWA 2004) that are good for vehicles but potentially unsafe for pedestrians. Faster turning speeds are discouraged by ODOT in the Highway Design Manual, and tighter angles are 4
encouraged instead (see Figure 2.1: Island geometry (ODOT)). Channelization has the potential to reduce the frequency of right-turn angle crashes (FHWA 2004). The Federal Highway Administration (FHWA) has tabulated the advantages and disadvantages of right-turn lanes (Table 2.1). Table 2.1: Potential Benefits and Liabilities of Right-Turn Lanes (FHWA 2015) CHARACTERISTIC POTENTIAL BENEFITS POTENTIAL LIABILITIES Separation of right-turn Higher speed of right-turning Safety vehicles from through traffic vehicles increases risk to pedestrians Higher right-turn capacity, Potential for off-tracking of large Operations shorter green time, less delay vehicles for following through vehicles, additional storage for approach queues None identified Longer pedestrian crossing distance, Multimodal time, and exposure; may require transit stop relocation None identified. Large intersection footprint Physical None identified Right-of-way/construction costs; Socioeconomic access restrictions to property None identified Periodic enforcement may be needed Enforcement, to prevent red light violations, education, and especially if right turns on red are maintenance prohibited Although channelization islands provide several benefits, a study conducted in Georgia on vehicle speeds at exclusive right-turn lanes concluded that treatments with right-turn lanes and raised islands resulted in the highest frequency of crashes among several treatment options (Fitzpatrick et al. 2006), including right-turn lanes with a lane line, island, shared through lane, or shared through lane with an island. In Texas, right-turn lanes with raised islands had the second highest amount of crashes (Fitzpatrick et al. 2006). Types of crashes included rear-end right-turn crashes, right angle and merge crashes, and sideswipe crashes (Fitzpatrick et al. 2006).
2.3
PHASING
The Traffic Signal Timing Manual defines a traffic signal phase as “a timing process, within the signal controller, that facilitates serving one or more movements at the same time” (USDOT 2015). Furthermore, “the MUTCD defines a signal phase as the right-of-way, yellow change, and red clearance intervals in a cycle that are assigned to an independent traffic movement or combination of traffic movements” (FHWA 2008). Generally, the National Electric Manufacturers Association phase numbering system combines right-turn and through movements because they are typically permitted (USDOT 2015). Figure 2.2 shows a typical phasing scheme for an intersection with permitted right-turn movements.
5
Figure 2.2: General phasing with permitted phasing for right turns (USDOT 2015) When the right-turn movement is a protected movement, overlap timing can be used (USDOT 2015). A sample phasing scheme is shown in Figure 2.3. Overlaps are most often used for rightturn movements when exclusive right-turn lanes exist (USDOT 2015). The parent phase is typically the compatible left-turn phase for right-turn overlaps (USDOT 2015). When the conflicting pedestrian phase is active, some traffic signal controllers allow omission of the rightturn overlap (USDOT 2015). This feature is used when the pedestrian phase is associated with the vehicular through movement (USDOT 2015). Both the compatible left-turn and adjacent through movements are parent phases for the right-turn overlap modifier feature (USDOT 2015). When the conflicting pedestrian phase is assigned to a modifier phase, the right-turn overlap is excluded only when there is a pedestrian call on the adjacent movement (USDOT 2015). When there is no pedestrian modifier, the right-turn overlap must run adjacent to the through movement as a permitted movement to avoid conflicts with pedestrians (USDOT 2015). Table 2 provides a diagram for typical right-turn overlap settings.
6
Figure 2.3: Phasing scheme for right-turn overlaps (USDOT 2015) Table 2.2: Typical Right-Turn Overlap Settings (FHWA 2015) PEDESTRIAN MODIFIER PHASE FOR MOVEMENT OVERLAP PARENT RIGHT-TURN OVERLAP OMIT (IF NUMBER LETTER1 PHASE AVAILABLE) 12 A 2* & 3 2P 14 B 4* & 5 4P 16 C 6* & 7 6P 18 D 8* & 1 8P 1
Agencies may have different overlap assignments based on their preference. * These phases should not be included as parent phases if a controller feature to omit right-turn overlap with active conflicting pedestrian phases is not available.
Furth et al. (2014) examined appropriate phasing for right turns and pedestrian/bicycle crossings under conditions of high turn volume or high speed, focusing on protected right-turn phasing. They introduced a unique “protected-yet-concurrent” phasing scheme. This scheme, which uses four rings rather than two, allows through movements to operate at a different time from the turning phase and at the same time as the parallel vehicular through phase. To illustrate the scheme and determine its likely effects on delay and street footprints, seven examples of concurrent phasing in the United States and the Netherlands were used. The study concluded that the delay and ROW requirements were minimal, and that the complexity of the phasing plans, coordination, and possibility of using re-service would affect phasing performance. Although this phasing scheme requires a right-turn lane, the authors argued that it uses time efficiently and is more efficient and economical than an exclusive pedestrian phasing scheme.
7
2.4
DISPLAY
Options to support right-turn signal phasing include permitted, protected-permitted, and protected displays (USDOT 2015). Right-turn signal installations and mode of operation are determined after an engineering study with consideration of factors related to capacity, congestion, and crash history (ODOT 2017). These factors include the presence of right-turn lane (s), right-turn volume and presence of conflicting crosswalk (ODOT 2017). If protected phasing is needed, then a separate overlap load switch must be provided for right-turn displays in exclusive right-turn lanes. This modification is preferred over combining the compatible left-turn signal phase with the right-turn arrow indication, as the latter tends to reduce flexibility in signal timing and effectiveness of traffic operations. Right-turn movements are permitted when an adjacent pedestrian phase is called (USDOT 2015). This permission can be signaled with a circular green indication. When no pedestrians are present, the right-turn movement is protected and can be signaled with a green arrow (USDOT 2015). The Signal Timing Manual states that “if a protected-only display is used with a pedestrian modifier function, the right-turn vehicular movement will be omitted when the conflicting pedestrian phase is called, and a right-turn red arrow will be displayed” (USDOT 2015). In Oregon, a right-turn movement that is faced with a circular red, or red arrow indication is permitted to turn after coming to a complete stop unless a posted sign states otherwise. This movement is called a permissive right-turn on red (ODOT 2015). Determination of right-turn signal phasing is based on engineering studies and factors including capacity, right-turn volume, and presence of congestion and related crashes, right-turn lane(s), and conflicting cross walk(s) (ODOT 2015). The permissive right-turn mode is the most commonly used and requires no signal head. Right-turn movements operate simultaneously with corresponding through movements but must yield to conflicting pedestrian movement (ODOT 2015).
Figure 2.4 Oregon signal head types for right-turns Figure 2.4 shows the signal types used in Oregon for right turns. Type 2 signal head is used for the permissive only movement, installed typically as a shared signal head and is applicable to the adjacent through movement also. Type 3R signal head is used for the true overlap. The not-ped feature is used if there is a conflicting crosswalk. Type 5 signal head is used for the “fauxoverlap”, where the ball indications are wired to the thru phases and the arrow indications are 8
wired to the complimentary left-turn phase. The green ball will be active with the parent thru phase green indication and the green arrow is active only when the complimentary left-turn parent phase is green. In the PPRT mode, the right-turn movement is protected during one part of the cycle and permissive during another (ODOT 2015). The protected portion generally occurs during the complementary left-turn phase, while the permissive portion occurs during the corresponding through-movement phase (ODOT 2015). The PPRT mode can provide operational benefits during heavy right-turn volumes (ODOT 2015). Figure 2.5 shows an example of PPRT phasing in a ring barrier diagram. The standard display NB right-turn phasing shown in Figure 2.5 is a Type 5 signal head. The arrow indications are wired to phase 1 and the circular indications are wired to phase 8.
Figure 2.5: Example scheme for PPRT phasing (ODOT 2015) Protected-only right-turns are generally used for exclusive right-turn lanes and can run concurrently with any non-conflicting vehicular or pedestrian movement (ODOT 2015). Traffic may only turn right when presented with a green arrow (ODOT 2015).
2.5
USE OF DETECTORS IN RIGHT-TURN OPERATIONS
Detectors used in right-turn lanes are sometimes designed to drop calls when permitted turns, such as right-turn-on-red (RTOR), are made and no other vehicles are present (USDOT 2015). A delay is placed on those detection zones to prevent calls from immediately being sent to the controller (USDOT 2015). These detectors are used to reduce delay by reducing calls due to RTOR and to call right-turn phases (USDOT 2015). The Traffic Signal Timing Manual defines delay as the temporary disabling of detector outputs for a phase (USDOT 2015). Delay values are determined by how long a vehicle is expected to occupy a detector before leaving it (USDOT 2015). Delay should be considered for right-lane detection when the capacity of the RTOR exceeds the right-turn volume or when a conflicting movement is on recall (USDOT 2015). However, delay may reduce the efficiency of the intersection if the RTOR capacity is limited (USDOT 2015). Typical delay settings range from 8 to 12 seconds (USDOT 2015), with larger delay values corresponding to higher cross-street volumes (USDOT 2015). 9
2.6
EFFECTS OF RESTRICTING RIGHT TURNS ON RED
Although developed as a fuel- and time-saving measure, RTORs can sometimes be detrimental for pedestrians (FHWA 2015a). Even though vehicle codes state that vehicles must come to a complete stop before they complete the RTOR, many drivers do not comply (FHWA 2015a). When drivers look to the right to check for approaching vehicles before they complete a turn, they tend to forget to check the crosswalk area for a crossing pedestrian. To gain more visibility, motorists tend to pull up into the crosswalk and block pedestrians from crossing (FHWA 2015a). When pedestrian volume is high, exclusive pedestrian phases and prohibition of RTORs should be considered, with signing used to inform motorists of the restriction (FHWA 2015a). The purpose of restricting RTORs is to increase pedestrian safety and reduce crash frequency at intersections (FHWA 2015a). RTORs may be prohibited specifically for operations at certain (e.g., busiest) times of the day (FHWA 2015a). Qureshi and Han (2001) determined that delay-related procedures at signalized intersections in the Highway Capacity Manual (HCM) do not effectively include effects of RTORs. They developed a queueing theory based on the use of queuing accumulation polygons that included RTORs. Two general queuing accumulation polygon patterns resulted from the four classes of signal phasing. Patterns consisted of two conflicting, two nonconflicting, and alternating conflicting and nonconflicting regimes. Delay was calculated from each pattern. The authors tested model performance using field and simulated data and determined that delays predicted by the models differed from observed and HCM delays. Based on several conditions and compared to the HCM approach, the proposed models predicted reductions in delay.
2.7
PEDESTRIANS
Exclusive pedestrian intervals have the potential to reduce the number of pedestrian crashes by 50% at sites of heavy pedestrian volume and low vehicle speed (FHWA 2015b). To give pedestrians a head start when entering an intersection at a crosswalk, a leading pedestrian interval may be used (NACTO 2012), which gives pedestrians a walk signal in advance of the green light for motorists (FHWA 2015b). Motorists are more likely to see and, therefore, to yield to pedestrians. To create self-enforcing yielding to pedestrians, pedestrian safety islands can be implemented. Channelized right-turn lanes can be eliminated to slow turning speeds (NACTO 2012). Tight corner radii, curb extensions, and pedestrian safety islands all force drivers to navigate intersections more cautiously (NACTO 2012). Pedestrians frequently are assigned to a through-traffic movement, with the assumption that a vehicle or bicyclist must yield before making a right turn (FHWA 2008). This phasing scheme may be inappropriate when exclusive right-turn lanes and high right-turn traffic volumes are present. In these cases, a lagging pedestrian interval may be used, in which the pedestrian walk interval does not start until a few seconds after the through-movement phase begins. This approach allows the waiting right-turn queue to clear before the pedestrian walk indication activates (FHWA 2008). Rao and Ni (2015) conducted a conflict analysis of two signalized intersections with different right-turn control schemes: one allowing RTORs and one with PPRTs. Trajectory information was collected at both intersections. A conflict risk assessment model was created based on the combination of collision probability and severity. “[A]ccording to signal schemes and lane function, cycle length and crosswalk were split into 16 several time and space intervals counted 10
as space-time cells, and pedestrian-vehicle conflict risk 17 degree of each space-time cell was calculated by risk assessment model” (Rao and Ni 2015). Conflicts between right-turn vehicles and pedestrians occurred more frequently with the RTOR scheme. Under the two signal control schemes, conflict risks for pedestrians and vehicles differed temporally, with highest risk during the pedestrian green-flash time with the RTOR scheme but during the pedestrian red time with the PPRT scheme. The study concluded that the PPRT vehicle control scheme can significantly reduce the number of conflicts between pedestrians and vehicles. The North Carolina Department of Transportation (NCDOT) developed a method to analyze design characteristics that affect pedestrians and bicyclists when crossing signalized intersections (Steinman and Hines 2004). Key features that were studied included crossing distance, roadway space allocation, corner radius dimensions, and traffic signal characteristics as they relate to level of service. The results have the potential to be used as a diagnostic tool to assess and improve pedestrian and bicyclist safety levels. Leden (2002) analyzed crash data from 300 intersections in Hamilton, Ontario, from 1983 through 1986. The study compared “pedestrian safety during semi-protected schemes, where left-turning vehicles face no opposing traffic but have potential conflicts with pedestrians, with pedestrian safety at normal non-channelized signalized approaches, where right-turning vehicles have potential conflicts with pedestrians.” Hourly flow for left- and right-turning vehicles were examined by four different approaches. Pedestrians encountered greater risk from left- than from right-turning vehicles. Risk and number of pedestrian accidents per pedestrian decreased with increased pedestrian flow, suggesting that increased pedestrian mode share can have a positive effect on pedestrian safety at signalized intersections. A study on pedestrian safety found that Florida had the highest rate of pedestrian fatalities in the United States during 2008–2011 (FDOT 2015). The study was conducted to understand driver behavior and compliance at signalized intersections when using four pedestrian signs: Stop Here on Red, No Turn on Red, Turning Vehicles Yield to Pedestrians, and Right on Red Arrow after Stop (FDOT 2015). This study applied two tools to improve efficiency of data processing: the Naturalistic Driving Study (NDS) Automatic Video Processing Tool, which detects and tracks pedestrians and traffic signal indications, and the NDS Data Reduction and Analysis Tool, which reviews and analyzes NDS videos and sensors (FDOT 2015). Data were divided into two groups for analysis: the feature group (including pedestrians) and the control group (excluding pedestrians) (FDOT 2015). Drivers showed the highest rate of compliance with the No Turn on Red sign (FDOT 2015). Stop on Red, No Turn on Red, and Right on Red Arrow after Stop features increased the likelihood of compliance compared to the control group (FDOT 2015). Compared to control sites, rates of compliance for features were higher when pedestrians were present (vs. not present) at the intersection (FDOT 2015). Pedestrian space and delay are used to measure the pedestrian’s level of service at signalized intersections. Hubbard et al. (2009) questioned whether this measure effectively reflects the negative effects that right-turning vehicles have on pedestrians. They used a binary logit model to identify factors affecting the likelihood that pedestrian safety will be compromised (i.e., the pedestrian is delayed or must alter their travel path or speed). Factors affecting pedestrian safety were the pedestrian direction of travel, right-turn volume, volume of pedestrians crossing,
11
whether the pedestrian arrived late, whether the pedestrian began crossing after the walk interval ended, and crosswalk characteristics.
2.8
BICYCLISTS
A recent study for ODOT was conducted to analyze right-hook (RH) crashes during the latter portion of the green phase at signalized intersections (Hurwitz, D. et al. 2015). A RH crash is an accident that occurs between a right-turning vehicle and through-moving bicycle. An extensive literature review of RH crashes was included in that document but is not repeated here. Experiments were jointly conducted by Oregon State University (OSU) and Portland State University (PSU) in the OSU Driving Simulator Lab (Jannat 2014). The study concluded that “78% of bicyclists were unaware of their stopping position with respect to stopped vehicles queued at an intersection during a red indication, and 19% of motorists reported that they would not yield to the adjacent bicyclist approaching from behind if they were detected in rear-view or side-view mirrors (Jannat 2014)”. When comparing crash and near-crash scenarios, the most common cause of failure was the driver’s failure to search actively for the bicyclist (Hurwitz, D. et al. 2015). Crash and near-crash situations were measured by time-to-collision. Several different intersection treatments (environmental factors, signage, curb radii, pavement markings, and protected intersection designs) were evaluated for their abilities to reduce the frequency and severity of RH crashes (Hurwitz, D. et al. 2015). Protected intersection designs performed well, with motorists being more likely to notice adjacent bicyclists compared to treatments that did not include this mitigation.
2.9
SAFETY
Analyzing the causes and rates of right-turn crashes at 68 right-turn approaches in Illinois, Schattler et al. (2015) sought to identify locations that could benefit from design modifications. Significantly higher crash and injury rates were correlated with approach head-turn angles greater than 140o, right-turn angles less than 45o, and intersection skew angles less than 75o (Schattler et al. 2015). The authors recommended several design modifications (still under investigation), such as decreasing the radius, removing or modifying channelized islands, adjusting the stop bar position, and providing pavement markings to separate paths for passenger cars when accommodating large trucks. Dixon et al. (2000) analyzed right-turn approaches in Georgia, with the aim of demonstrating the limitations of certain right-turn configurations and identifying mitigation options. Finding limited research on right-turn treatments at signalized intersections, the authors divided common right-turn treatments into entrance and exit treatments (Figure 2.6Figure 2.6). “Entrances,” used to describe the upstream, were defined as entry points and geometric features used by vehicles to perform right turns. “Exits” were defined as geometric conditions or traffic control devices where right-turning vehicles entered conflicting traffic streams. Geometric conditions and traffic control devices are dependent on each other and were evaluated together (Dixon et al. 2000). When traffic islands were used, fewer right-angle crashes were observed. Use of an exclusive turn lane correlated with an increase in the number of sideswipe crashes. In the absence of additional controls, the use of an exclusive right-turn lane on the cross street did not reduce the number of rear-end crashes. These findings could serve as a basis for further research of factors such as traffic volume, conflicts, road types, and limited right-turn treatments. 12
Figure 2.6: Entrance (left) and exit (right) treatments (Dixon et al. 2000) Conflicts at signalized intersections and reductions in safety can result from improper geometric design and signal settings (Chai and Wong 2014). A study conducted at Nanyang Technological University simulated four cross-intersection scenarios using a cellular automata model (Chai and Wong 2014). Relationships between conflicts involving right-turn vehicles, traffic volume, and right-turn movement control strategies were analyzed (Chai and Wong 2014). The authors estimated traffic volume impacts, permissive right-turn vs. red-amber-green arrow, shared vs. right-turn lanes, and signal settings (Chai and Wong 2014). They concluded that prediction models can be developed on the occurrence and severity of vehicle conflicts based on different geometric layouts and control strategies (Chai and Wong 2014).
2.10 SUMMARY The planning, design, and operation of signalized intersections are complex processes that require the balancing of safety and efficiency for all system users. Right-turn-lane geometry is an important variable affecting safety and operational efficiency. Pedestrians must be accommodated for safe operation. While the flashing yellow arrow is allowed for PPRT operations, there has been no previous research on the operational or safety implications of PPRT phasing.
13
14
3 ANALYSIS OF CRASH DATA Right-turn crash data at signalized intersections in Oregon were reviewed for years 2011–2013 (Table 3.1). Crash records were filtered to identify all crashes that occurred at signalized intersections, 1 which accounted for 18–20% of reported crashes in 2011–2013. Crashes at signalized intersections were filtered to identify those involving a right-turning vehicle 2. For 2011–2013, an average of 848 crashes per year involved right-turning vehicles at signalized intersections (1.7% of all crashes, 8.8% of crashes at signalized intersections). Table 3.2 decomposes data for right-turn crashes at signalized intersections by crash severity. Nine fatal crashes, 1,146 injury crashes, and 1,389 property-damage-only crashes were associated with right-turn movements at signalized intersections in Oregon during 2011–2013. Table 3.1: Summary of ODOT Crash Data (2011–2013) CATEGORY 2011 2012 Total reported crashes Crashes at traffic signals Right-turn crashes at traffic signals
49,053
49,798
9,183 (18.7%)
9,690 (19.5%)
809 (1.7%)
881 (1.8%)
2013 49,510 10,061 (20.3%) 854 (1.7%)
3-YEAR AVG 49,454 9,645 (19.5%) 848 (1.7%)
Table 3.2 Severity of ODOT Right-Turn Crashes at Signalized Intersections (2011–2013) CRASH SEVERITY 2011 2012 2013 3-YEAR AVG 3 (0.4%) 3 (0.3%) 3 (0.4%) 3 (0.4%) Fatal 377 (46.6%) 423 (48.0%) 346 (40.5%) 382 (45.0%) Injury 455 (51.7%) 505 (59.1%) 463 (54.6%) Property damage only 429 (53.0%) 809 881 854 848 Total The researchers sought comparable published tabulations for comparison. The National Highway Traffic Safety Administration (NHTSA) analyzed causes of crashes (Choi 2010), using data collected at crash scenes during 2005–2007. Among crashes at intersections, 52.5% occurred at intersections with at least one traffic signal, and 1.8% of vehicles at these signalized intersections were turning right before the crash (see Appendix A). The NHTSA did not find any statistically significant and critical reasons of driver error for the right-turn crashes. Reasons for differences in percentages of right-turn crash contributions in Oregon vs. nationally are likely related to differences in included crash types. No other analysis of statewide-signalized intersection data by turning movement was found in the literature. Next, characteristics of right-turn crashes in Oregon were compared to characteristics of all other crashes at signalized intersections. Proportions of right-turn and intersection crashes were 1 2
TRAF_CNTL_DEVICE_SHORT_DESC = “TRF SIGNAL MVMNT_SHORT_DESC = “TURN-R”
15
compared by a proportions test, with the null hypothesis assuming no difference based on collision type. Null (Eq. 3.1) and alternate (Eq. 3.2) hypotheses are listed below: 𝑯𝑯𝟎𝟎 : 𝒑𝒑𝟏𝟏 = 𝒑𝒑𝟐𝟐
(3.1)
𝑯𝑯𝑨𝑨 : 𝒑𝒑𝟏𝟏 ≠ 𝒑𝒑𝟐𝟐
(3.2)
The z-value was computed by the following equations. 𝒁𝒁 =
�− 𝒑𝒑𝟐𝟐 ) (𝒑𝒑 𝟏𝟏 � 𝟏𝟏
𝟏𝟏
� (𝟏𝟏− 𝒑𝒑 �)( + ) �𝒑𝒑 𝒏𝒏𝟏𝟏 𝒏𝒏𝟐𝟐 𝒚𝒚
𝒚𝒚
�𝟏𝟏 = 𝟏𝟏 , � 𝒑𝒑 𝒑𝒑𝟐𝟐 = 𝒏𝒏𝟐𝟐 𝒏𝒏 𝟏𝟏
�= 𝒑𝒑
𝒚𝒚𝟏𝟏 +𝒚𝒚𝟐𝟐
𝒏𝒏𝟏𝟏 +𝒏𝒏𝟐𝟐
𝟐𝟐
(3.3)
(3.4) (3.5)
Where y 1 and y 2 are the number of right-turn and intersection crashes, respectively, and n 1 and n 2 are the total number of right-turn and intersection crashes for 2011–2013. Statistically significant differences (P ≤ 0.05; gray in Table 3.3) were found between proportions of right-turn vs. intersection crashes for all collision types except head-on and parking maneuver collisions. The turning movement difference would be expected because the right-turning crash would likely be typed as a “turning movement” (the most significant event is used to type the crash). There were small proportions of backing, sideswipe (meeting and overtaking), miscellaneous, and noncollision crashes, such that statistically significant differences are not necessarily critical. Large differences in pedestrian and fixed-object collisions may indicate a potential area for simulator studies. As expected, rear-end and angle crashes were less frequent.
16
Table 3.3: Comparison of Proportions of Crashes at Intersections by Collision Type COLLISION TYPE RIGHTALL RIGHTALL P TURN OTHER TURN % OTHER CRASHES CRASHES % 1 287 0.04% 1.07% 0.00 Backing * 4 10 0.04% 0.01 Miscellaneous 0.16% * 218 625 2.32% 0.00 Pedestrian 8.57% * 37 5,034 1.45% 18.69% 0.00 Angle * 2 46 0.08% 0.17% 0.27 Head-on 184 12,976 7.23% 48.18% 0.00 Rear-end * 0 47 0.00% 0.17% 0.03 Sideswipe-meeting * 22 465 0.86% 1.73% 0.00 Sideswipe-overtaking * 1,905 6,819 25.32% 0.00 Turning movement 74.88% * 4 17 0.16% 0.06% 0.09 Parking maneuver 11 37 0.14% 0.00 Non-collision 0.43% * 156 571 2.12% 0.00 Fixed-object or other6.13% * object 2,544 26,934 Total * Statistically significant at 95% confidence level; Bold text indicates differences where proportions were significantly higher for right-turn crashes.
Table 3.4 shows differences in crash proportions by cause of crash and results of statistical analysis (with similar table formatting applied). Statistically significant differences were observed for crashes due to not yielding ROW (to pedestrians, bicyclists, or other vehicles), passing the stop sign or red flasher, improper overtaking, following too close, making improper turns, other non-specified reasons (not improper driving), improper change of traffic lanes, driver drowsy/fatigued, disregarding traffic control device, presence of non-motorist illegally in the roadway, non-motorist clothing not visible, load shifting on a vehicle, and inattention/careless driving. Of these, the greatest differences in proportions were observed for crashes caused by following too closely, making an improper turn, and not yielding ROW.
17
Table 3.4: Comparison of Proportions of Crash Types at Intersections by Cause of Crash CRASH CAUSE RIGHTALL RIGHT ALL P TURN OTHER -TURN OTHER CRASHES CRASHES % % 176 710 6.92% 2.64% 0.80 Speed too fast for conditions 885 4411 16.38% 0.00* 34.79% Did not yield right of way 2 22 0.08% 0.08% 0.00* Passed stop sign or red flasher 227 6213 8.92% 23.07% 0.96 Disregarded R-A-G traffic signal 9 51 0.35% 0.19% 0.08 Drove left of center on twoway road 73 83 0.31% 0.00* 2.87% Improper overtaking 256 10835 10.06% 40.23% 0.00* Followed too closely 590 816 3.03% 0.00* 23.19% Made improper turn 77 818 3.03% 3.04% 0.98 Other improper driving 0 11 0.00% 0.04% 0.31 Mechanical defect 11 40 0.15% 0.00* 0.43% Other (not improper driving) 20 381 0.79% 1.41% 0.01* Improper change of traffic lanes 22 129 0.48% 0.01* 0.86% Disregarded other traffic control device 7 42 0.28% 0.16% 0.16 Wrong way on one-way roadway 81 0.08% 0.30% 0.04* Driver drowsy/fatigued/sleepy 2 37 111 0.41% 0.00* 1.45% Nonmotorist illegally in roadway 4 11 0.04% 0.01* 0.16% Nonmotorist clothing not visible 0 5 0.00% 0.02% 0.49 Vehicle improperly parked 2 0.00% 0.01% 0.66 Defective steering mechanism 0 6 58 0.24% 0.22% 0.83 Inadequate or no brakes 2 1 0.00% 0.00* 0.08% Vehicle lost load or load shifted 1 0.00% 0.00% 0.76 Tire failure 39 0.04% 0.14% 0.17 Phantom/non-contact vehicle 1 34 851 1.34% 3.16% 0.00* Inattention 3 46 0.12% 0.17% 0.53 Driving in excess of posted speed 1 1 0.00% 0.04* 0.04% Speed racing 42 687 1.65% 2.55% 0.01* Careless driving 49 405 1.93% 1.50% 0.10 Reckless driving 1 6 0.04% 0.02% 0.59 (blank) 2,544 26,934 Total * Statistically significant at 95% confidence level
18
3.1
SUMMARY
A review of Oregon crash data for 2011–2013 revealed that 18–20% of all crashes annually occurred at signalized intersections, of which 1.7–1.8% were attributed to right-turning movement (accounting for 8.8% of all crashes at signalized intersections). Most of these crashes (54.6%) were property-damage-only crashes. Comparison of right-turn and intersection crash proportions revealed large, statistically significant, and relevant differences in angle, pedestrian, rear-end, and fixed-object crash types. Following too closely, not yielding ROW, and improper turns were crash causes with significantly different proportions between right-turn and all intersection crashes.
19
20
4 SURVEY A survey was conducted to understand how well Oregon drivers comprehend different right-turn signal display alternatives. Open-ended and multiple-choice questions were developed to elicit each driver’s understanding and self-reported response to typical traffic signal displays for controlling right turns. The survey was conducted online, with recruitment through postcards that were mailed to 9,872 addresses in Oregon. This chapter describes the development and administration of the survey and the results of the analysis.
4.1 DESIGN AND REFINEMENT The first step in designing the survey was the development of a generic template for survey images. The research team designed the initial image template by considering a recent Florida DOT report (Boot et al. 2015). A Google Sketch Up image was used instead of a real photo, to enable explicit modification of the scene. Every effort was made to present questions neutrally, allowing respondents to provide meaningful answers reflecting their comprehension of the signal indications. Past questions on other surveys on flashing yellow arrow comprehension were used as a guide. Several rounds of review and refinement followed the internal development of the survey questions. Transportation graduate students at OSU and PSU and members of the project Technical Advisory Committee (TAC) from ODOT tested a pilot survey and provided feedback for further improvements of the format and content of the survey questions. Once the project team was satisfied with the survey design, the survey was finalized. The finalized survey, distribution methods, and record handling were reviewed and approved by the IRB of PSU (163752 IR).
4.2 INSTRUMENT The survey consisted of 21 questions. The first section of the survey included open-ended questions, which asked respondents to report their understanding of right-turn signal display indications. Specific questions regarded comprehension of the green ball, green arrow, red ball, red arrow, and flashing yellow arrow indications. For these questions, respondents were presented a computer image of an intersection from a driver’s perspective and were instructed to assume that they were turning right. The scale of the signal heads was slightly enlarged to make the displays more prominent in the image. Survey images are shown in the results and analysis section for each question. Two versions of each question were developed: one version with and one version without a Right Turn Only (RTO) sign. The survey was designed such that half of the respondents were randomly administered one version with the RTO sign and the other half were administered the version without the sign for all open-ended questions. In the second section of the survey, respondents were given a set of multiple-choice questions. They were asked to provide their reasoning for what they perceived as either similarities or differences between the red ball and red arrow and between the green ball and flashing yellow arrow signal indications. The third and final section of the survey consisted of closed-ended 21
multiple-choice demographic questions on the respondent’s income and education levels, driving habits, and eyesight. The entire survey instrument is included in Appendix B.
4.3 ADMINISTRATION A survey response rate of 6–8% was assumed based on a previously conducted postcard/online design by researchers at PSU (Currans et al., 2015). To generate sufficient responses for analysis, a sample size of 10,000 respondents was selected. A sampling scheme was designed based on the proportion of the population in each county (Table 4.1). Based on this scheme, a random sample of addresses within each county was purchased through Info USA. After removing incorrect/ missing addresses from the purchased address sample, there remained 9,872 households to which recruitment materials could be sent. Each household was assigned a unique ID number.
22
Table 4.1: Sampling Scheme for Driver Survey REGION COUNTY POPULATION 1
2
3
4
5
Clackamas Hood River Multnomah Washington Clatsop Columbia Tillamook Yamhill Polk Marion Lincoln Linn Benton Lane Douglas Curry Coos Josephine Jackson Wasco Sherman Gilliam Jefferson Wheeler Crook Deschutes Lake Klamath Morrow Umatilla Union Wallowa Baker Grant Harney Malheur Oregon
397,385 24,245 777,490 570,510 37,750 50,390 25,690 103,630 78,570 329,770 47,225 120,860 90,005 362,150 109,910 22,470 62,990 83,720 210,975 26,370 1,790 1,975 22,445 1,445 21,085 170,740 8,010 67,110 11,630 79,155 26,625 7,100 16,425 7,430 7,295 31,480 4,013,845
POPULATION % 22.46 1.37 43.94 32.24 3.03 4.04 2.06 8.32 6.31 26.47 3.79 9.70 7.22 29.06 22.43 4.59 12.85 17.08 43.05 8.22 0.56 0.62 6.99 0.45 6.57 53.20 2.50 20.91 6.21 42.30 14.23 3.79 8.78 3.97 3.90 16.82
23
# OF ADDRESSES 990 60 1937 1421 94 126 64 258 196 822 118 301 224 902 274 56 157 209 526 66 4 5 56 4 53 425 20 167 29 197 66 18 41 19 18 78 10,000
A recruitment postcard (shown in Figure 4.1) containing pertinent information about the survey objectives and the online link was sent to each address. On the other side of the postcard, the household ID number was included, which respondents were required to enter while answering the survey. Survey responses were never linked to names of respondents answering the survey, thus ensuring confidentiality of responses. Recipients were provided with the option of providing their contact information at the end of the online survey, to be entered into a drawing for one of five $100 Amazon.com gift cards.
Figure 4.1: Recruitment postcard for the survey
4.4 RESPONSE RATE Postcards were mailed to 9,872 addresses. A total of 416 respondents clicked the online link to begin the survey; however, only 399 respondents completed the survey. No postcards were returned as undeliverable. The calculated response rate was 4%. Figure 4.2 shows the geographic distribution of respondents. Responses were received from all five ODOT regions. The lower response rate for this survey compared to previous surveys using similar methods was attributed to two reasons. First, the shortened URL that was used for the online link on the postcard was case-sensitive. Typing the online link exactly as presented on the postcard led to the correct survey (bit.ly/DriverStudy), but typing all lowercase letters (bit.ly/driverstudy) sent recipients to a paper at the University of Wisconsin on autonomous vehicles. The research team fielded calls and emails from some recipients and clarified the link; however, it is possible that some recipients typed the link in lowercase letters and did not contact the research team for clarification when the link did not work. Second, due to the project schedule, response postcards were mailed on 16 May 2016 from the FedEx processing center in California. The first postcards arrived in Oregon a week before Memorial Day weekend. Initial responses were strong, and then a drop off in response rate was observed over the holiday weekend. 24
Figure 4.2: Geographic distribution of respondents
4.5 DEMOGRAPHIC SUMMARY Of the 399 people who responded to the survey, 397 people provided some or all of the requested demographic information. Table 4.2 presents demographic information for all survey respondents, categorized by whether they received the version of the survey with or without the RTO sign (“Sign” vs. “No Sign” in tables). Percentages for Oregon from the Census Bureau are shown where available. Older, educated white males were overrepresented as survey respondents compared to 2010 Census estimates for Oregon (US Census). Survey respondents were 61% male compared to 49% male for the total population. Survey respondents were generally older than the general population, with larger representation in the 55–64 and 65+ years categories. Survey respondents were 93% White/Caucasian (vs. 79% reported in the Census). According to the US Census American Community Survey (ACS) data, approximately 30% of Oregonians have a Bachelor’s degree or higher, compared to over 65% of respondents in the sample. The ACS reports that 89.5% of Oregon residents have a high-school education or higher, compared to 98% of sample respondents. About 71% of survey respondents reported household incomes of less than $100,000, which compares well to the Census data of 75%. Proportions did not differ between respondents who took the survey with or without the RTO sign.
25
Table 4.2 Demographic Comparison between Survey and Census Respondents CATEGORY DEMOGRAPHIC SURVEY SURVEY % SURVEY % CENSUS VARIABLE % (SIGN) (NO SIGN) % Male 60.7 59.3 62.1 49.2 Gender Female 39.3 40.7 37.9 50.8 (n = 397) 18–24 2.0 1.5 2.5 Age 25–34 8.3 11.2 5.4 13.7 (n = 399)
Race (n = 375)
Income (n = 336)
Education (n = 396)
35–44 45–54 55–64 65+ American Indian or Alaska Native Asian Black or African American Hispanic or Latino/a White or Caucasian Other Less than $25,000 $25,000 – $50,000 $50,000 – $75,000 $75,000 – $100,000 $100,000 – $150,000 $150,000 – $200,000 $200,000 or more No schooling completed, or < 1 year Nursery, kindergarten, and elementary grades (1–8) High school (grades 9–12, no degree) High-school graduate (or equivalent) Some college (1–4 years, no degree) Associate degree (incl. occup. or academic degrees) Bachelor’s degree (BA, BS, AB etc.) Master’s degree (MA, MS, MENG, MSW, etc.) Professional school degree (MD, DDC, JD etc.) Doctorate degree (PhD, EdD, etc.)
15.3 14.5 29.3 30.6 0.5
16.8 12.2 28.6 29.6 1.1
13.8 16.7 30.0 31.5 0.0
13.1 14.1 13.3 13.8 1.1
2.1 0.5 2.4 92.5 1.9 9.2 19.0 21.4 21.1 19.6 6.3 3.3 0.0
1.1 0.0 3.3 91.8 2.7 6.6 18.0 24.6 18.6 21.0 7.2 4.2 0.0
3.1 1.0 1.6 93.2 1.0 11.8 20.1 18.3 23.7 18.3 5.3 2.4 0.0
3.6 1.7 11.7 78.5 3.3 23.6 23.2 17.0 11.5 13.4 5.7 5.6
0.0
0.0
0.0
2.0
1.5
2.5
6.5
5.8
5.7
5.9
24.5
18.7
16.5
20.8
26.6
11.1
11.3
10.9
8.2
33.3
38.1
28.7
18.9
19.4
18.0
20.8
11.2
4.8
6.2
3.5
4.8
2.6
6.9
4.1
26
Table 4.3 summarizes the driving behaviors of respondents. Respondents tended to drive multiple times in a week (97%), to be licensed for over 10 years (96%), and to hold an Oregon driver’s license (98%). More than half (58%) of respondents reported that they drove more than 10,000 miles each year. A small percentage of respondents (3%) indicated that they were colorblind. Most respondents indicated that they used corrective glasses or contacts for vision (65%). Table 4.3: Characteristics of Survey Respondents CATEGORY DEMOGRAPHIC SURVEY % VARIABLE Less than 1 time per 2.0 week 1 time per week 0.8 Driving 2 – 4 times per week 15.0 frequency 5 – 10 times per week 32.1 More than 10 times per 50.1 week 1 – 2 years 0.5 3 – 5 years 1.5 Driver’s license 6 – 10 years 2.3 10+ years 95.7 Less than 5,000 14.3 5,000 – 9,999 27.8 Miles driven per 10,000 – 14,999 30.3 year 15,000 – 19,999 16.8 Greater than 20,000 10.8 97.7 Oregon driver’s Yes license No 2.3 Yes 2.5 No 96.5 Color blind Don’t want to provide 1.0 this information/Don’t Know Yes 65.0 No 34.0 Corrective glasses or Don’t want to provide 1.0 contacts this information/Don’t Know
SURVEY % SURVEY % (SIGN) (NO SIGN) 2.0 2.0 0.5 14.3 33.7 49.5
1.0 15.8 30.5 50.7
0.0 1.0 2.6 96.4 12.8 28.6 30.1 15.8 12.8 97.4 2.6 3.1 95.9 1.0
1.0 2.0 2.0 95.0 15.8 27.1 30.5 17.7 8.9 98.0 2.0 2.0 97.0 1.0
65.3 34.7 0.0
65.0 34.0 1.0
4.6 RESULTS AND ANALYSIS 4.6.1 Coding Because the survey contained open-ended questions designed to assess comprehension of various signal display indications, the responses needed to be categorized for further analysis. The 27
research team reviewed each open-ended response. Responses were coded as correct, partially correct, or incorrect by two reviewers independently, based on established criteria shown in Table 4.4. The same coding convention was followed for coding both the responses from the survey and the actions observed in the simulator experiments, therefore providing a basis for comparison of the results from both experiments. Table 4.4: Error Coding of Narrative DISPLAY CORRECT INDICATION Turn right with Steady Circular caution after yielding Green to pedestrians in the crosswalk
Turn right without stopping recognizing that the steady green arrow indication means a protected movement (or) Indicated that they would watch for pedestrians who may cross against the pedestrian Don’t Walk signal Steady Circular Red Come to a complete stop and complete the and Steady Red turn when they found Arrow a safe gap or remained stopped if they failed to find a gap Turn right with Flashing Yellow caution after yielding Arrow to pedestrians in crosswalk Steady Green Arrow
PARTIALLY CORRECT Turn right without stopping but failed to state that they would yield to pedestrians if present in the crosswalk Check for pedestrians and turn right (or) slow down and check for pedestrians and other cross traffic but did not recognize the protected movement in either case
INCORRECT Stop before turning
Stop before turning
Stop or turn right, without providing additional details
Stop and remained stopped until the green indication
Turn right without stopping or failed to state that they would yield to pedestrians if present in the crosswalk
Stop before turning
For the steady circular green, responses were coded as correct if the respondents indicated that they would turn right with caution after yielding to pedestrians in the crosswalk. A response was coded as incorrect, if the respondents indicated that they would stop before turning. Partially 28
correct codes were assigned to responses when the respondents stated that they would turn right without stopping, but failed to mention yielding to pedestrians if present, in the crosswalk. A correct response for the steady green arrow was coded when the respondent recognized that it was a protected movement and stated that they would turn right without stopping. A response was also coded as correct, if the respondent stated that they would watch for pedestrians who may jaywalk against the signal indication. A partially correct response was coded when the respondents stated that they would check for pedestrians and turn right or if they slowed down to check for pedestrians and other cross traffic, but did not give any indication that they recognized that a steady green arrow implied a protected movement. Responses were coded as incorrect, when the respondents stated that they would stop and turn right. The coding convention followed for the steady circular red and the steady red arrow indications was the same, as according to Oregon law the expected correct action for both display indications is the same for right turns. A response for these indications was coded as correct, when the respondents stated that they would come to a complete stop and complete the turn after finding a safe gap or they would stop and remained stopped, if they failed to find a suitable gap. A partially correct response was coded when the respondents stated that they would stop or turn right, without providing additional details. A response was coded as incorrect, if the respondents stated that they would stop and remained stopped until the green indication. For the flashing yellow arrow signal display indication, responses were coded as correct, if the respondents stated that they would turn right with caution after yielding to pedestrians in the crosswalk. A partially correct response was coded, when respondents stated that they would turn right without stopping or failed to state that they would yield to pedestrians if present in the crosswalk. A stated response of stopping before turning was coded as incorrect. Interrater reliability was assessed by Cohen’s kappa coefficient ĸ, a statistic that measures interrater agreement for categorical items and is calculated as follows: 𝐏𝐏𝐏𝐏(𝒂𝒂)−𝐏𝐏𝐏𝐏(𝒆𝒆)
𝜿𝜿 =
𝟏𝟏−𝑷𝑷𝑷𝑷(𝒆𝒆)
(4-1)
where Pr(a) represents the actual observed agreement, and Pr(e) represents the chance agreement. Pr(e) is calculated using the following formula: 𝐏𝐏𝐏𝐏(𝒆𝒆) =
𝒄𝒄𝒄𝒄𝟏𝟏 ∗𝒓𝒓𝒓𝒓𝟏𝟏 𝒄𝒄𝒄𝒄𝟐𝟐 ∗𝒓𝒓𝒓𝒓𝟐𝟐 �+( ) 𝒏𝒏 𝒏𝒏
�
𝒏𝒏
(4-2)
where cm1 is the column 1 total, cm2 is the column 2 total, rm1 is the row 1 total, rm2 represents the row 2 total, and n is the number of observations. This statistic can range between -1 and +1, where 0 represents agreement due to random chance, and 1 represents perfect agreement between the raters (McHugh 2012). Kappa statistic values of 0.61–0.80 indicate substantial agreement, and those of 0.81–1.00 indicate almost perfect agreement. Cohen’s kappa statistic was calculated for the steady circular green, steady green arrow, steady circular red, steady red arrow, and flashing yellow arrow questions separately with and without the RTO sign. For all but one question, one independent coding trial was conducted. For the green arrow question, two coding trials were conducted. Following estimation of the kappa
29
statistic (trials 1 and 2), phone meetings between the research team members were conducted, and coding discrepancies were resolved. Kappa values are shown in Table 4.5. Table 4.5: Estimated Values of Cohen’s Kappa Coefficient CATEGORY TRIAL 1 TRIAL 1 TRIAL 2 TRIAL 2 (SIGN) (NO (SIGN) (NO SIGN) SIGN) 0.86 0.88 1.00 1.00 Green Ball 0.77 0.65 0.75 0.74 Green Arrow 0.79 0.84 1.00 1.00 Red Ball 0.89 0.91 1.00 1.00 Red Arrow 0.81 1.00 1.00 Flashing Yellow 0.86 Arrow
TRIAL 3 TRIAL 3 (SIGN) (NO SIGN) 1.00
1.00
4.6.2 Open-ended Comprehension Questions Each respondent was asked five open-ended questions to determine their comprehension of the steady red circular ball, steady red arrow, steady green circular ball, steady green arrow, and flashing yellow arrow signal indications for right-turn movements. Respondents were presented with this wording for each of the five displays: Imagine that you are approaching the intersection in the lane farthest to the right and planning to TURN RIGHT. What action would you take based on the current signal display? Please type your response in the box below and be as descriptive as possible. Responses to each question were reviewed and classified as correct, partially correct, or incorrect. Coding criteria for each display are provided in each subsection. A sample of the open-ended responses is included in Appendix C. A discussion of each of these signal display indications follows. 4.6.2.1 Steady Circular Green Ball Participants were presented the image shown in Figure 4.3 and asked to imagine that they were driving in the right lane with a steady circular green ball indication. They were prompted to describe their resulting course of action. Responses were coded as following the coding convention outlined in Table 4.4. Results of the analysis of the responses are shown in Table 4.6. Most respondents (73%, n=398) correctly indicated that they would turn right and yield to pedestrians in the crosswalk. The remaining respondents largely indicated that they knew that they had the ROW to proceed (25%) but did not include any descriptions of yielding to pedestrians before turning so were coded partially correct. Only 2.3% stated that they would stop before turning and were coded incorrect. Groups with or without the RTO sign differed in the proportion of drivers who stated that they would yield to pedestrians (70%, n=195 with vs. 76%, n=205 without the sign). These differences were not statistically significant based on the two-proportion z test (p-value = 0.14 for correct, 0.10 for partially correct, and 0.78 for incorrect responses). 30
Figure 4.3: Image used for open-ended question on steady green circular ball signal indication (with RTO sign) Table 4.6: Responses to Open-ended Question on Steady Green Circular Ball Signal Indication RESPONSE OVERALL (N = 398) SIGN (N = 195) NO SIGN (N = 203) 73.1% 69.7% 76.4% Correct 24.6% 28.2% 21.1% Partially Correct 2.3% 2.1% 2.5% Incorrect 4.6.2.2 Steady Green Arrow Respondents were presented with the image shown in Figure 4.4 and asked to describe their course of action when faced with a steady green arrow signal indication in the rightturn lane. Responses were coded following the coding convention outlined in Table 4.4. Table 4.7 summarizes the results for this question, which was answered by 397 participants (195 with and 202 without the RTO sign). The exclusive movement communicated by the green arrow was not well understood by the respondents. Only 63% (n=397) of respondents correctly stated that the steady green arrow communicates a completely protected movement and that they did not need to yield to pedestrians and other vehicles. However, 32.7% were coded as partially incorrect because they indicated that they would check for pedestrians and turn right, or if they indicated that they would slow down and check for pedestrians and other cross traffic, but did not recognize the exclusive protected movement. This is a fail-safe comprehension response and many respondents preferred to be cautious and check for pedestrians before turning. Respondents presented with the image including the RTO sign (n=195) responded correctly at a rate 11% higher than respondents who were not presented the sign (n=202). The results were statistically significantly different in proportions of correct responses (pvalue = 0.03 for correct, 0.06 for partially correct, and 0.47 for incorrect responses)..
31
Figure 4.4: Image used for open-ended question on steady green arrow signal indication (with RTO sign) Table 4.7: Responses to Open-ended Question on Steady Green Arrow Signal Indication RESPONSE OVERALL (N = 397) SIGN (N = 195) NO SIGN (N = 202) 63.5% 68.7% 58.4% Correct 32.7% 28.2% 37.1% Partially Correct 3.8% 3.1% 4.5% Incorrect 4.6.2.3 Steady Circular Red Participants were provided the image in Figure 4.5 and asked a question designed to assess their comprehension of the steady circular red signal indication for right-turn movement. Responses were following the coding convention outlined in Table 4.4. Table 4.8 summarizes the findings for this question, which was answered by 398 participants (195 with and 203 without the RTO sign). The steady circular red signal indication was well understood by respondents with 83% (n=398) of respondents stating that they would come to complete stop and complete the turn when they found a safe gap, or that they would remained stopped if they failed to find a gap. An additional 7% were coded partially correct because they did not provide additional detail other than stopping. Proportions of correct, partially correct, and incorrect responses were similar between groups with (n=195) and without (n=203) the RTO sign (p-value = 0.26 for correct, 0.20 for partially correct, and 0.76 for incorrect responses).
32
Figure 4.5: Image used for open-ended question on steady circular red signal indication (with RTO sign) Table 4.8: Responses to Open-ended Question on Steady Circular Red Signal Indication RESPONSE OVERALL (N = 398) SIGN (N = 195) NO SIGN (N = 203) 83.2% 81.0% 85.2% Correct 7.0% 8.7% 5.4% Partially Correct 9.8% 10.3% 9.4% Incorrect 4.6.2.4 Steady Red Arrow Participants were provided the image shown in Figure 4.6 and asked a question designed to ascertain their comprehension of the steady red arrow signal indication. Responses were coded following the coding convention outlined in Table 4.4. Table 4.9 summarizes results for this question, which was answered by 398 respondents (195 with and 202 without the RTO sign). This display had the lowest comprehension with only 52% (n=398) stating that they would come to a complete stop and would turn if an acceptable gap was available after yielding to other traffic and pedestrians or would remain stopped if they failed to find a gap (Oregon law allows RTOR after coming to a complete stop regardless of red ball or red arrow). An additional 7% were coded partially incorrect (they did not provide additional details). The remaining 41% were coded as incorrect and the most common incorrect response was the perception that drivers needed to remain stopped until the indication changed to green. The proportions of incorrect responses were statistically significant different in proportions, 34% (n=195) with and 47% (n=202) without the sign, p-value = 0.01 for correct, 0.77 for partially correct, and 0.02 for incorrect responses.
33
Figure 4.6: Image used for open-ended question on steady red arrow signal indication (with RTO sign) Table 4.9: Responses to Open-ended Question on Steady Red Arrow Signal Indication RESPONSE OVERALL (N = 397) SIGN (N = 195) NO SIGN (N = 202) 52.1% 57.9% 46.5% Correct 7.3% 7.7% 6.9% Partially Correct 40.6% 34.4% 46.5% Incorrect 4.6.2.5 Flashing Yellow Arrow Respondents were presented the image shown in Figure 4.7 and asked a question designed to probe their comprehension of the flashing yellow arrow indication. Online, the yellow arrow flashed at the appropriate interval (an animated GIF image was created). Responses were coded following the coding convention outlined in Table 4.4.. Table 4.10 presents the results for this question, which was answered by 398 respondents (195 with and 203 without the RTO sign). Most respondents (77%, n= 398) understood the expected driver response from the FYA indication and stated that they would exhibit caution while turning and yield to pedestrians and cross traffic while coming to a stop if necessary. An additional 3.8% were coded partially incorrect because although the respondents indicated that that they would make the turn with caution while yielding they did not mention stopping. The remaining 19.6% were coded incorrect stated that they would come to a complete stop before turning or would remain stopped until the signal display became green. The group provided the RTO sign had a significantly higher proportion of correct responses than the group that was not provided this sign (81% vs. 72%, p-value = 0.04). Groups with (n=195) and without (n=203) the RTO sign did not differ significantly in their proportions of partially correct (p-value = 0.22) and incorrect responses (p-value = 0.12). 34
Figure 4.7: Image used for open-ended question on flashing yellow arrow signal indication (with RTO sign) Table 4.10: Responses to Open-ended Question on Flashing Yellow Arrow Signal Indication RESPONSE OVERALL (N = 398) SIGN (N = 195) NO SIGN (N = 203) 76.6% 81.0% 72.4% Correct 2.6% 4.9% Partially Correct 3.8% 19.6% 16.4% 22.7% Incorrect
4.6.3 Multiple-Choice Comprehension Questions Four sets of close-ended multiple-choice comprehension questions were posed to respondents regarding signal displays for right-turn movements. The first two sets of questions concerned participants’ comprehension of the steady red arrow and flashing yellow arrow indications and their level of confidence in their responses. The third and fourth sets concerned participants’ perceptions of the similarities and differences between the red arrow and circular red indications, and between the circular green and flashing yellow arrow indications, for right-turning movements. Each of these question sets is discussed further below. 4.6.3.1 Steady Red Arrow Comprehension Respondents were presented an image of a steady red arrow with the following wording to gauge steady red arrow comprehension: In Oregon, if you are turning right and you see the steady red right arrow display, what would be the appropriate response? How confident are you of your answer? Three choices were presented to respondents – they can turn right cautiously without stopping, they can turn right but first must come to a complete stop and find a gap before turning (correct response), or they must stop and wait until they receive a green 35
indication before turning. Table 4.11 shows the results from the survey. Overall, 53% of respondents chose the correct response. The group presented the RTO sign had a higher proportion of correct responses, although differences between groups were not significant (with sign n = 196, without sign n = 203; p-value = 0.12 for correct or incorrect responses). Regardless of whether the sign was displayed, many respondents did not comprehend the correct action to be taken when faced with a steady red arrow signal indication. As shown in Table 4.12, respondents were generally very confident (56%) or confident (30%) in their responses, with no significant difference in proportions between groups with and without the RTO sign. These results, coupled with the comprehension results, show some confusion among Oregon drivers regarding the steady red arrow and the appropriate course of action. Table 4.11: Responses to Multiple-choice Question on Steady Red Arrow Signal Indication RESPONSE OVERALL SIGN NO SIGN (N = 399) (N = 196) (N = 203) 0% 0% Turn right cautiously without stopping 0% 52.6% 56.6% 48.8% Complete stop and find a gap before turning 47.4% 43.4% 51.2% Stop and wait for a green indication before turning Table 4.12: Confidence Levels for Responses to Multiple-choice Question on Steady Red Arrow Signal Indication CONFIDENCE LEVEL OVERALL (N = 399) SIGN (N = 196) NO SIGN (N = 202) 56.3% 56.1% 56.4% Very confident 29.9% 30.6% 29.2% Confident 5.0% 5.1% 5.0% Neutral 6.5% 6.6% 6.4% Somewhat confident 2.3% 1.5% 3.0% Not at all confident 4.6.3.2 Red Arrow and Circular Red Comparison Respondents were asked if the red arrow and circular red signal indications for the rightturn movement were the same or different as a two-choice question. Signal display indications are shown in Figure 4.8. Respondents were asked to explain the reasoning behind their choice. Table 4.13 shows the results from the survey for this question.
36
Figure 4.8: Image used for question on comprehension of red arrow vs. circular red signal indications Table 4.13: Comprehension of Red Arrow Vs. Circular Red Signal Indications SAMPLE SIMILAR DIFFERENT 49.6% 50.4% Overall (n = 395) 51.5% 48.5% Sign (n = 194) 47.8% 52.2% No sign (n = 201) Overall, 395 respondents answered this question (with sign n = 194, without sign n = 201). Half of respondents thought that the circular red and red arrow signal displays were the same, and half thought that they were different. No significant differences in comprehension or between groups with and without the RTO sign were found (p-value = 0.45 for both same and different). Among respondents who indicated that the signal displays were different, the overwhelming reasoning for the difference was the perception that with the red ball, drivers were supposed to stop and were allowed to proceed if they were able to find a gap before turning; whereas with the red arrow, they indicated that drivers needed to come to a complete stop and remain stopped until they get a green indication. Incorrect/missing responses regarding the correct course of action for the right-turn arrow display in Oregon mirrored the findings discussed earlier with the rightturn arrow open-ended question. 4.6.3.3 Flashing Yellow Arrow Comprehension Respondents were presented with the following wording to gauge their understanding of the flashing yellow arrow indication and their confidence in their response: In Oregon, if you are turning right and you see the flashing yellow arrow display, what would be the appropriate response? How confident are you of your answer? Table 4.14 shows the responses obtained from the survey. Overall, 76% of respondents chose the correct response – turn cautiously right without stopping. A higher proportion of correct responses was found among respondents who were presented with the RTO sign for previous questions, but the difference in proportions was not statistically 37
different between groups. Stopping and waiting for a green indication before turning was the only option that yielded statistically significant differences in proportions with and without the sign (with sign n = 196, without sign n = 203; P = 0.19 for correct, 0.38 or 0.05 for incorrect responses). However, this difference was the result of the very small sample size. As shown in Table 4.15, 86% of respondents were confident or very confident in their responses, with similar confidence levels between groups with and without the RTO sign. Table 4.14: Responses to Multiple-choice Question on Flashing Yellow Arrow Indication RESPONSE OVERALL SIGN NO SIGN (N = 399) (N = 196) (N = 203) 75.7% 78.6% 72.9% Turn right cautiously without stopping 23.3% 21.4% 25.1% Complete stop and find a gap before turning 1.0% 0.0% 2.0% Stop and wait for a green indication before turning Table 4.15: Confidence Level for Responses to Multiple-choice Question on Flashing Yellow Arrow Indication CONFIDENCE LEVEL OVERALL SIGN NO SIGN (N = 398) (N = 195) (N = 203) 50.0% 48.5% 51.5% Very confident 35.7% 37.2% 34.2% Confident 5.5% 6.1% 5.0% Neutral 5.8% 4.6% 6.9% Somewhat confident 3.0% 3.1% 3.0% Not at all confident 4.6.3.4 Flashing Yellow Arrow, Circular Green Comparison Respondents were asked to indicate their perceptions of similarities and/or differences between the flashing yellow arrow and the steady circular green signal indications, along with the reasoning behind their choice. A total of 395 responses were received for this question, with the results summarized in Table 4.16. An overwhelming majority of respondents (91%) indicated that the two signal displays meant different things to them as drivers, while only 9% indicated that they were the same. Examination of the openended responses revealed that most people associated green displays with proceeding with the movement and having ROW, but associated flashing yellow displays with the need to slow down, exercise caution, and yield if necessary. Of the respondents who indicated that the indications were the same, their reasoning was that both indications require the driver to yield to pedestrians in the crosswalk before proceeding. Among the groups with and without the RTO sign, the proportions were similarly distributed across the two categories, with most respondents stating that the flashing yellow arrow and the steady circular green signal indications were different. No statistically significant differences in proportions were observed with and without the RTO sign (with sign n = 194, without sign n = 201; p-value = 0.77 for same or different). 38
Figure 4.9: Image used for question on comprehension of flashing yellow arrow vs. circular green signal indications Table 4.16: Comprehension of Flashing Yellow Arrow Vs. Circular Green Signal Indications SAMPLE SIMILAR DIFFERENT 9.4% 90.6% Overall (n = 395) 9.8% 90.2% With sign (n = 194) 91.0% Without sign (n = 201) 9.0%
4.7 SUMMARY A survey was conducted to understand how well Oregon drivers comprehend different right-turn signal display alternatives. The survey was conducted online with recruitment through postcards that were mailed to 9,872 addresses in Oregon. The survey consisted of 21 questions. The first section of the survey included open-ended questions which asked respondents to report their understanding of green ball, green arrow, red ball, red arrow, and flashing yellow arrow indications. Respondents were presented an image of intersection and asked “Imagine that you are approaching the intersection in the lane farthest to the right and planning to TURN RIGHT. What action would you take based on the current signal display”? Respondents were randomly presented with intersections with or without a “Right Turn Only (RTO)” sign placed overhead on the signal mast arm adjacent to the signal display. In the second section of the survey, respondents were given a set of multiple-choice questions to provide their reasoning for what they perceived as either similarities or differences between the red ball and red arrow and between the green ball and flashing yellow arrow signal indications. The third and final section of the survey consisted of closed-ended multiple-choice demographic questions on the respondent’s income and education levels, driving habits, and eyesight. Overall, the survey received responses from a wide geographical area of Oregon but was over-represented by older white males compared to statewide demographic estimates from Census data.
39
The research team reviewed each open-ended response and coded them as correct, partially correct, or incorrect by two reviewers independently, based on established criteria for each signal display. Interrater reliability was assessed by Cohen’s kappa coefficient ĸ, a statistic that measures interrater agreement for categorical items. Initial coding revealed substantial agreement for most displays except the green arrow. After a series of iterations and discussions, all coding inconsistencies were resolved. Overall, analysis of the of 399 completed survey open-ended responses revealed that most respondents understood the steady circular green (73% correct), steady green arrow (63% correct), steady circular red (83% correct). Even though the FYA display for right turns is uncommon in Oregon and most drivers surveyed are not likely to have encountered the display, comprehension was high (77% correct). In the multiple-choice response, 75% correctly identified the response “Turn right cautiously without stopping” as the expected response. There was a general misunderstanding of the required driver response for the steady red arrow signal indication in the open-ended responses (52% correct). These were confirmed in the follow-up multiple-choice questions, where nearly 50% of the respondents incorrectly and confidently (86%) identified the correct response to the steady red arrow to be “stop and wait for a green indication before turning”. With respect to the steady red arrow, there are a number of possible explanations for this misunderstanding. First, most drivers see red arrows for left turns, where the correct response is almost always to stop and wait for the light to turn green. Second, driver laws pertaining to the correct course of action for red arrow signal display indications also vary significantly between states (Oregon and California are completely opposite). Interestingly, for the flashing yellow arrow and steady red arrow displays, higher proportions of correct responses were observed when the RTO sign was present versus when this sign was absent. Other than the sign clearly indicating the signal head is for right-turn movements, it provides no additional guidance about the expected response. The prior research on the flashing yellow arrow signal indication did not find significant differences in comprehension with the use of additional signage.
40
5 MICROSIMULATION A signalized intersection with PPRT phasing was constructed in a microsimulation environment. The objective was to explore the operational performance of several PPRT phasing alternatives under varying volumes (right-turn vehicle volumes, conflicting pedestrian movements, and conflicting left-turn vehicular movements). The software VISSIM was used to build the simulated intersection. Within VISSIM, the ASC/3 software-in-the-loop signal controller was used for signal timing and phasing purposes. The following sections describe the simulation configuration and results.
5.1 SIMULATION MODEL BACKGROUND Numerous tools – analytical and simulation models – have been developed for analyzing traffic. Analytical models use mathematical formulations to determine traffic states (capacity, density, speed, delay, and queuing) of small-scale facilities (Akçelik 2007). Multimodal in nature, simulation models are used to model traffic flows and interactions between modes in a transportation network. Both tools are useful in evaluating design alternatives and making decisions. Simulation models are categorized as macroscopic, mesoscopic, or microscopic. In macroscopic models, the simulation takes place on a section basis, without explicitly considering individual vehicles. Some well-known examples of macroscopic simulation models are PASSER, SYNCHRO, TRANSYT, and TRANSYT7F. Mesoscopic models are a blend of macroscopic and microscopic models. Microscopic models model the movement of individual vehicles in the traffic stream based on car-following and lane-changing models. Popular microsimulation models are PARAMICS, AIMSUN, VISSIM, SIMTRAFFIC, and CORSIM. Microsimulation models are increasingly being used as analysis tools worldwide and offer several advantages, such as an ability to model system wide effects of alternatives and various geometric configurations. Although these models can provide detailed statistics, they often require large amounts of data, the accuracy of which affects the precision of results. Microsimulation models need to be properly calibrated and validated to yield accurate results. Some degree of user skill is required to build a representative model. Each of the physical aspects, whether hardware or software, must be created within the microsimulation model itself. These hardware/software-in-the-loop (HITL/SITL) systems are the mathematical models of the hardware/software that are used to represent the hardware/software within the simulation itself.
5.2 TYPES OF SIMULATION MODELS Traffic flows at signalized intersections can be modeled by HITL or SITL systems. HITL systems require a simulation model on the computer which interfaces with a controller interface device and an actual traffic signal controller. The controller interface device passes the detector calls and phase status between the traffic controller and the microsimulation model. Figure 5.1 shows the data flow in a HILS system. Various simulation modeling packages, such as CORSIM, SimTraffic and VISSIM, are capable of supporting HITL applications. Major 41
disadvantages with HILS systems are the inability to speed up or slow down the simulation, the lack of synchronization between controllers and computer clocks, and the need for separate controller hardware for each intersection (Urbanik et al. 2006).
Figure 5.1: HILS data flow (Stevanovic et al. 2009) SITL systems use a microscopic simulation model, a virtual traffic controller that runs on the same computer, and an interface that allows for communication and exchange of information between the microscopic simulation and the virtual traffic controller (Stevanovic et al. 2009). The biggest advantage of SILS is the ability to speed up or slow down the simulation. Two SITL applications have been developed – Siemens’s NextPhase, which connects to CORSIM and VISSIM, and Econolite’s ASC/3, which connects to VISSIM (Stevanovic et al. 2009). The ASC/3 virtual controller is the only commercially available SITL system. The ASC/3 virtual controller embedded in VISSIM was developed by PTV America and Econolite Control Products in cooperation with the University of Idaho. The ASC/3 virtual controller software in VISSIM is identical to the software that runs in controllers in the field. Figure 5.2 shows the data flow in a SITL model. In this study, the ASC/3 virtual traffic controller was used in VISSIM to simulate a signalized intersection with PPRT phasing.
Figure 5.2: Data flow in SITL model (Stevanovic et al. 2009)
5.3 MODEL DEVELOPMENT Steps undertaken for model development were selecting the site and gathering input data (geometry, traffic volumes, vehicle composition, and signal timing parameters). These steps are described in detail in the following sections.
42
5.3.1 Site Selection Selection of the site was primarily based on discussions with TAC members, which revealed that PPRT phasing using a FYA display was present at two Oregon intersections – NW Cornell Rd. and NW Evergreen Pkwy. in Beaverton and NW 3rd St. and NW Van Buren Ave. in Corvallis, Oregon. The research team chose to simulate the intersection of NW Cornell Rd. and NW Evergreen Pkwy. in VISSIM because that location was perceived to be more transferable due to the presence of a one-way approach at the Corvallis location. Figure 5.3 shows the simulation site, which was located to the west of downtown Portland.
Figure 5.3: Site location for simulation
5.3.2 Geometry At this intersection (Figure 5.4), Cornell Rd. in the eastbound (EB) direction has one through lane, one shared through/right lane, and an exclusive left-turn lane. The westbound (WB) approach has two through lanes, exclusive right- and left-turn lanes, and a bike lane to the left of the exclusive right-turn lane. The southbound (SB) approach has exclusive left- and right-turn lanes, a shared left/through lane, and a curb bike lane. The northbound (NB) approach is minor and is one lane that is shared among all movements. All of the lane lengths and widths used in the simulation model were measured by using Bing satellite images present in VISSIM.
43
Figure 5.4: Intersection geometry
Figure 5.5: Volumes used in the simulation model 44
5.3.3 Volumes Volumes used in the VISSIM simulation were provided to the research team in the form of a Synchro file by Washington County, Oregon, the agency responsible for operating the signal. The research team selected the PM peak period volume data (Figure 5.5), because turning movement counts were readily available for this period.
5.3.4 Vehicle Composition Vehicle composition was coded as 98% passenger cars and 2% heavy vehicles, based on classification data obtained from Washington County. For passenger cars, each vehicle was assumed to be of single occupancy. Default driving behaviors within VISSIM were used for both vehicle type.
5.3.5 Signal Timing Signal timing plans for this intersection were obtained from Washington County. This intersection operates as a 6-phase intersection, with protected/permitted left- and right-turn movements. The major street left-turn movements (phases 1 and 5) are protected/permitted with a flashing yellow arrow signal display. Additionally, PPRT phasing is enabled for phase 6 (WB to NB movement). Split phasing is present at this intersection for side street phases 4 and 8 as seen in Figure 5.6. Typically, this intersection is coordinated with cycle lengths ranging from 100 to 130 seconds, depending on the time of day. However, this intersection was operated as an isolated intersection, while retaining the same signal timing parameters (min green, max green, gap times, etc.), because the research team felt that coordination issues were not directly relevant to permitted right-turn performance.
Figure 5.6: Phase rotation diagram for NW Cornell Rd. and NW Evergreen Pkwy.
45
For the WB right-turn movement, which has PPRT phasing, three indications are displayed. A green arrow is displayed when the non-conflicting SB left turn is active. A flashing yellow arrow is displayed when the concurrent WB green is active. If a pedestrian is present on phase 6 and a call is received by the controller, then a red arrow indication is shown for the WB right-turn movement. The research team invested substantial time in exploring how to make the flashing yellow indication work simultaneously within the ASC/3 virtual controller and VISSIM for the left- and right-turn permitted movements at this intersection. The team used VISSIM 7.0 with overlaps to allow the left- and right-turn flashing yellow arrows to function appropriately.
5.4 RESULTS 5.4.1 Base Model Initial base model runs were conducted by using the volumes shown in Figure 5.5. Ten runs were simulated for the base model, and average delays per movement and average green times were extracted. Figure 5.7 shows the average delays per movement derived from the base case model.
Figure 5.7: Average delays by movement Across all movements, the average control delay was 15 seconds per vehicle. The WB right turn, which is the movement of interest, experienced average delays of less than 4 seconds. This short delay time was probably related to the low volume of pedestrians (10 per crosswalk per hour) used in the simulation. Both the WB and WB left-turn movements (which were protected/permitted) showed small delays. Longest delays (50–60 seconds) were observed for the SB movements and were probably due to the small average green time of 20 seconds (Table 5.1). As expected, major street phases 2 and 6 had the highest average green times. Walk times for pedestrian phases ranged between 5 and 7 seconds. 46
Table 5.1 Average Green Times From Simulation PHASE BASE 1 4.3 2 72.1 4 19.8 5 5.0 6 69.7 8 5.5 Ped 2 5.0 Ped 4 6.0 Ped 6 7.0
5.4.2 Sensitivity Analysis To examine how volume affected delays, a sensitivity analysis was undertaken. Six different scenarios were tested: compared to the base case, (1) EB left-turn volumes (phase 5) were increased by 10%, (2) EB left-turn volumes (phase 5) were increased by 25%, (3) WB right-turn volumes (phase 6) were increased by 10%, (4) WB right-turn volumes (phase 6) were increased by 25%, (5) pedestrian volumes on phase 6 were increased by 5 times, and (6) pedestrian volumes on phase 6 were increased by 10 times. Table 5.2 shows the percent difference in delays for the six scenarios, compared to the base case. For scenarios 1 and 2, increasing volumes for the EB left turn did not result in significant delay increases for any phase. Minor increases in delays were observed for the WB through phase, possibly due to the flashing yellow arrow indication. Increasing the right-turn vehicular volumes in scenarios 3 and 4 resulted in increased delays for the WB right-turn movement. Increases in pedestrian volumes in scenarios 5 and 6 had the greatest effects in terms of delay increases for both WB right and EB left turns. This result is intuitive because both of these movements conflict with pedestrian movements. Overall, delays showed little to no change with an increase in EB left- or WB right-turn volumes. The largest delays were the result of the increase in pedestrian volumes.
47
Table 5.2 Percent Change in Delay, by Movement and Phase % DIFF TO 10% EB 25% EB 10% WB 25% WB BASE LT LT RT RT ---SB LT (TH/LT) -----SB TH -1% --SB RT ----SB LT 5% 4% --4% WB TH 2% 1% 11% 7% WB LT 2% 18% 41% WB RT (PPRT) -1% 3% 2% -EB TH -12% -2-2% -1% EB RT 2% -3% 2% 4% EB LT ----2% NB TH 1% 1% -1% -1% NB RT ----NB LT --5% 4% Ped 02 EB ---1% 1% Ped 02 WB ----Ped 04 SB ----Ped 04 NB 1--4% -6% -6% Ped 06, WB --6% 1% 1% Ped 06, EB 1% 1% --1% All
50 PEDS --6% ---49% --5% -------21% 2% 7%
100 PEDS --13% ---74% --13% -------107% 79% 27%
-- indicates no change (0%)
5.5 SUMMARY This chapter describes the development of a microsimulation model and the resulting analyses. The objective of the microsimulation was to explore the operational performance of PPRT phasing alternatives under varying volumes (right-turn pedestrian movements and conflicting left-turn vehicular movements) to provide guidance for interpreting results from the driving simulator. A base model, based on a real-world intersection, was simulated in VISSIM. Sensitivity analysis was performed by increasing conflicting left-turn, right-turn, and pedestrian volumes in six different scenarios. As expected, results from the simulation models indicate that the pedestrian volumes had the greatest effects on delays.
48
6 DRIVING SIMULATOR EXPERIMENT This chapter describes an experiment conducted in the OSU Driving Simulator Laboratory. The experiment was designed to use surrogate safety measures (driver performance and decision making) to evaluate the performance of PPRT phasing alternatives.
6.1 EXPERIMENTAL EQUIPMENT The experimental design and established experimental protocols were selected to address the research questions of interest. This approach is grounded in accepted practice (Fisher et al. 2011) and leverages the unique research capabilities at OSU. Two primary tools were used for this experiment, the OSU driving simulator and the Applied Science Laboratories (ASL) eye-tracking system, which are described in detail in the following sections.
6.1.1 Driving Simulator The OSU driving simulator facility consists of two primary components: a desktop development simulator and a full-scale high-fidelity motion-based simulator. Researchers first built and test drove the environment using the desktop development simulator. The multimonitor platform of the desktop development simulator (Figure 6.1), with the incorporated steering wheel and floor pedals, is used for creating, coding, and testing developed scenes. This desktop development simulator allows for quick troubleshooting during environment development.
Figure 6.1: Operator workstation for driving simulator. Left: Designing an experiment in the Internet Scene Assembler with Javascript. Right: Researcher evaluating a newly designed environment The full-scale OSU driving simulator is a high-fidelity motion-based simulator comprising a full 2009 Ford Fusion cab mounted above an electric pitch motion system capable of rotating ±4°. 49
The vehicle cab is mounted on the pitch motion system, with the driver's eye point located at the center of the viewing volume. The pitch motion system allows for the accurate representation of acceleration or deceleration (Swake et al. 2013). Three liquid crystals on silicon projectors with a resolution of 1,400 × 1,050 are used to project a front view of 180° × 40°. These front screens measure 11 ft × 7.5 ft. A digital light-processing projector is used to display a rear image for the driver’s center mirror. The two side mirrors have embedded LCD displays. The update rate for projected graphics is 60 Hz. Ambient sounds around and internal sounds in the vehicle are modeled with a surround-sound system. The computer system includes a quad-core host running Realtime Technologies SimCreator Software (version 3.2) with a 60-Hz graphics update rate. The simulator software is capable of capturing and outputting accurate values for performance measures (speed, position, brake, and acceleration). Figure 6.2 shows views of the simulated environment created for this experiment from inside (left) and outside (right) the vehicle.
Figure 6.2: Simulated environment in OSU driving simulator, from the participant’s perspective inside (left) and from outside (right) the vehicle The full-scale driving simulator is controlled from the operator workstation (Figure 6.3). The full driving simulator is located in a separate room from the desktop development simulator and the full simulator operator workstation. This separation prevents participants in the vehicle from being affected by visual or audible events from researchers during the experiment.
50
Figure 6.3: Operator workstation for full-scale driving simulator. Monitors are shown displaying SimObserver (left), and the simulated environment (center) or vehicle dashboard (right) as seen in the vehicle The virtual environment was developed by using Simulator software packages, including Internet Scene Assembler (ISA) (version 2.0), SimCreator, and Blender (version 2.45). The simulated test track was developed in ISA by using Java Script-based sensors that change the signal indication and display dynamic objects, such as pedestrians crossing the street towards the turning vehicle when the participant vehicle approaches. 6.1.1.1 Simulator Data The following parameters describing the participant vehicle were recorded at roughly 60 Hz (60 times per second) throughout the duration of the experiment: •
Time – Maps changes in the speed, acceleration, and position of the participant vehicle relative to the location of the intersection;
•
Instantaneous speed of participant vehicle – Identifies changes in speed when the driver approaches an intersection;
•
Instantaneous position of participant vehicle – Estimates the headway and distance upstream from the stop line of the participant vehicle when approaching an intersection;
•
SimObserver data – The driving simulator is equipped with five cameras positioned at various viewing angles to observe the actions of participants when approaching an intersection. Figure 6.4 shows the various camera views and screen captures that were recorded by SimObserver (version 2.02.4). 51
Figure 6.4: Screenshots of the six views from SimObserver. Clockwise from top left: simulated scene as projected on the screen; driver’s upper body and hands on steering wheel; acceleration and brake pedals in vehicle; entire simulator viewed from outside the vehicle; steering wheel and dashboard; and driver’s face. 6.1.1.2 Simulator Sickness Simulator sickness is a phenomenon wherein a person exhibits symptoms similar to motion sickness due to use of a simulator (Fisher et al. 2011; Owens and Tyrrell 1999). Symptoms can include headache, nausea, dizziness, sweating, and in extreme situations, vomiting. Although there is no definitive explanation for simulator sickness, one widely accepted theory, cue conflict theory, suggests that it arises from the mismatch of visual motion cues and physical motion cues, as perceived by the vestibular system (Owens and Tyrrell 1999).
6.1.2 Eye Tracker In conjunction with the driving simulator, an eye-tracking system was used to record where participants were looking while driving in the simulator. Eye-tracking data were collected with the ASL Mobile Eye-XG platform (Figure 6.5), which allows the user unconstrained eye and head movements. A 30-Hz sampling rate was used, with an accuracy of 0.5–1.0° (OSU Driving and Bicycle Research Lab 2011). Gaze was calculated based on the correlation between the participant’s pupil position and the reflection of three infrared lights on the eyeball. Eye movement consists of fixations and saccades. Fixations occur when the gaze is directed towards a particular location and remains still for some period of time (Green 2007; Fisher et al. 2011). Saccades occur when the eye moves between fixations. The ASL Mobile Eye-XG system records a fixation when the participant’s eyes pause in a certain position for more than 100 milliseconds. Quick movements to another position (saccades) are calculated indirectly from the dwell time 52
between fixations. Total dwell times are recorded by the equipment as the sum of the time of fixations and saccades consecutively recorded within an area of interest (AOI).
Figure 6.5: OSU researcher demonstrating the Mobile Eye XG Glasses (left) and Mobile Recording Unit (right)
6.2 EXPERIMENTAL DESIGN To address research questions related to Oregon drivers’ comprehension of different right-turn signal display alternatives, an experiment was designed by using the OSU driving simulator and the eye-tracker equipment. Two types of signal heads and five different indications were created for use in the simulator scenarios. Additionally, a pedestrian signal head indication (with either the walking person or upraised hand background) was added to each scenario. Signal heads were developed by using Blender (version 2.71). Figure 6.6 provides examples of signal heads as seen in the simulator from the perspective of an approaching driver. Ped Signal
Signal Heads
Figure 6.6: Example of signal and pedestrian head configurations in the simulator environment
53
6.2.1 Road and Intersection Geometry Intersection approaches included one through lane and an exclusive right-turn lane, along with a single receiving lane for the right-turn movement. The posted speed limit was 35 mph. The roadside was light-to-medium-density commercial and industrial development, and light ambient traffic was included. The roadway cross-section consisted of two 12-ft traffic lanes in each direction with no median (Figure 6.7), while the cross-section of the roadway receiving the rightturn roadway consisted of two 10-ft traffic lanes in each direction with no median (Figure 6.8). A yellow centerline, solid white edge line, small 1-ft paved shoulder, and 6.5-ft-wide pedestrian sidewalks on both sides of the road were constantly present.
Figure 6.7: Screenshot of a sample environment coded in the simulator
Figure 6.8: Screenshot of a cross-section of the right-turn roadway environment coded in the simulator
54
6.2.2 Experimental Variables 6.2.2.1 Independent Variables Three independent variables were included in the experiment: signal indication type and active display, length of the right-turn bay, and presence of pedestrians (Figures 6.9, 6.10, and 6.11). These variables were selected by the research team in collaboration with the ODOT Technical Advisory Committee (TAC) to help answer the research questions. The first independent variable, “signal indication type,” had two and three levels for each type of signal head: circular red and circular green, and solid red arrow, solid green arrow, and flashing yellow arrow. The flashing yellow arrow had two levels: walk interval on (FYAW) and clearance interval on (FYAC). Signal displays and arrangements were as realistic as possible, and no additional signs (e.g., Right Turn Only) were present. The second independent variable, “right-turn bay length,” had two levels: 50 and 100 ft. The length of the turning bay was determined based on an examination and statistical analysis of the lengths of turning bays on 21 Oregon State highway intersections. The final independent variable, “presence of a conflicting pedestrian in the crosswalk,” had two levels: 1) no (zero) pedestrians present and 2) one conflicting pedestrian walking towards the participant’s vehicle.
Figure 6.9: Example scenario of a participant approaching the right turn on a red arrow
55
Figure 6.10: Example scenario of right turns with 50- and 100-ft turning bay lengths
Figure 6.11: Example scenario of right turns with and without the presence of pedestrians 6.2.2.2 Dependent Variables Three primary dependent variables were observed based on the research questions and independent variables selected for this experiment. Visual attention was recorded from the eye-tracking equipment as participants glanced towards AOIs (defined later). Driver decision making was observed by driver behavior (stop, yield, or go) in response to the signal display and phasing. The speed of the participant vehicle was observed from the simulator data to determine how participants accelerate/decelerate while approaching a right-turning bay. These changes can demonstrate potentially unsafe driver behavior, such as sharp braking or crossing a lane line into conflicting traffic.
6.2.3 Factorial Design A factorial design was chosen for this experiment to enable exploration of all three independent variables separately. The factorial design for the three variables, each with two or three levels, resulted in the inclusion of 24 scenarios, which were presented within subjects. The within56
subject design provides advantages of greater statistical power and reduced error variance associated with individual differences (Cobb 1998). However, one fundamental disadvantage of the within-subject design is the existence of “practice effects,” caused by practice, experience, and growing familiarity with procedures as participants move through the sequence of conditions. To control for practice effects, the order of the presentation of scenarios to participants needs to be randomized or counterbalanced (Girden 1992). Table 6.1 summarizes the independent variables and their associated levels in the factorial design. Table 6.1: Experimental Variables and Levels VARIABLE ACRONYM CATEGORY LEVEL 1 SHA 2 1 Nominal 2 Signal Head (categorical) SHB 3 Discrete Geometry
G
Pedestrians
P
1 2
Discrete
1 2
LEVEL DESCRIPTION CR: Circular Red CG: Circular Green SRA: Solid Red Arrow SGA: Solid Green Arrow W: Walk FYA: interval Flashing Yellow C: Clearance Arrow walk interval TB1: Right-turn bay length 1: 50 ft TB2: Right-turn bay length 2: 100 ft No pedestrians crossing Pedestrians crossing
6.2.4 Counterbalancing and Presentation of Driving Scenarios To control for the practice or carryover effect, the order of the scenarios was counterbalanced. Six different track layouts were developed and presented in random order to each participant. Randomized, partial counterbalancing was chosen due to its simplicity and flexibility in terms of statistical analysis and number of required participants. Each track had four right-turning maneuvers, and each right turn was randomly assigned one level for each of the three independent variables. To provide more variability in track presentation, the start and finish locations of these grids were varied. Right-turning scenarios were interrupted by through and left-turn movements at intersections that were not experimental scenarios to prevent participants from anticipating the motivation of the study and to reduce the risk of simulator sickness. Table 6.2 presents the configuration layout for each of the 24 right-turn scenarios that were presented to participants, in a randomized order, across six tracks (labeled as T # in the table heading). Figures 6.9 to 6.11 show examples of individual scenarios in the simulator as presented to the drivers. Each right turn had one level for each of the three independent variables. Figure 6.12 shows an example grid layout with four right-turning scenarios (Grid 1). The “Path” in the 57
Figure indicates the sequence of intersections through which the participant drove. Layouts of other grids with two, three, and four right-turning scenarios are included in Appendix D. Participants were instructed to turn right at an intersection through an automated voice command saying “Please, turn right at the next intersection.” This voice command was generated by a Java Script-based sensor placed at the right-turning intersection approach, which was triggered by the presence of the participant vehicle on the sensor. The command was set to announce twice at 400- and 200-ft in advance of the right-turn movement. For left-turn and through movements, the participant heard automated voice commands saying “Please, turn left at the next intersection” and “Please, go straight,” respectively. Finally, participants were instructed to stop the vehicle at the end of each track through an automated voice command saying “Please, stop the vehicle.” Table 6.2: Grids and Right-turning Intersection Layout T#
RT #
SIGNAL HEAD
6 22 9 14
1 2 3 4
FYAC SGA SRA CG
23 8 2 18
1 2 3 4
FYAW CG CG FYAC
Grid 2 100 100 50 50
Pedestrian crossing No pedestrian crossing No pedestrian crossing Pedestrian crossing
Grid 3 100 50 100 50
Pedestrian crossing No pedestrian crossing No pedestrian crossing No pedestrian crossing
19 5 7 4
1 2 3 4
CR FYAW CR SGA
20 12 10 21
1 2 3 4
CG FYAC SGA SRA
16 1 11 17
1 2 3 4
SGA CR FYAW FYA
3 13 15 24
1 2 3 4
SRA CR SRA FYAC
BAY LENGTH (ft) Grid 1 50 100 100 50
Grid 4 100 100 100 100 Grid 5 50 50 100 50 Grid 6 50 50 50 100
58
PEDESTRIAN No pedestrian crossing Pedestrian crossing No pedestrian crossing Pedestrian crossing
Pedestrian crossing No pedestrian crossing No pedestrian crossing Pedestrian crossing Pedestrian crossing No pedestrian crossing No pedestrian crossing Pedestrian crossing No pedestrian crossing Pedestrian crossing Pedestrian crossing Pedestrian crossing
Figure 6.12: Example layout of Grid 1 with four right-turning (RT) scenarios – Path StartRight-Right-Right-Through-Right-Right-Left-Left-Right-Finish
6.3 DRIVING SIMULATOR EXPERIMENTAL PROTOCOL The experimental procedure was carefully designed to reduce the occurrence of simulator sickness, such as by providing long tangential sections between intersections or small breaks between driving successive grids. The entire data collection process was designed to ensure that all necessary information was recorded efficiently. This section describes the step-by-step procedures of the driving simulator study, as conducted for each individual participant.
6.3.1 Recruitment A total of 52 individuals, primarily from the community surrounding Corvallis, Oregon, were participants in the experiment. The population of interest was licensed drivers; therefore, only licensed drivers residing in Oregon with at least 1 year of driving experience were recruited for 59
the experiment. Participants were required to not wear glasses (contacts were acceptable), to be physically and mentally capable of legally operating a vehicle, and to be deemed competent to provide written, informed consent. Participants were recruited through flyers posted around campus and the surrounding community and through emails sent to different campus organizations and email listservs. Older participants were specifically recruited by email using the Center for Healthy Aging Research registry (LIFE Registry), which includes people 50 years or older who reside in Oregon and wish to volunteer for research studies. Researchers did not initially screen interested participants based on gender; however, once the quota for men was reached, only women were allowed to participate. Although it was expected that many participants would be OSU students, an effort was made to incorporate participants of all ages within the specified range of 18 to 75 years. Throughout the entire study, participant data were kept under double-locked security in compliance with accepted Institutional Review Board (IRB) procedures (Study #7435). Each participant was randomly assigned a number to remove any uniquely identifiable information from the recorded data.
6.3.2 Informed Consent and Compensation When the test participant arrived at the laboratory, they received the OSU IRB-approved informed consent document, which described the reasoning behind the study, the importance of participation, and the risks and benefits of the test for the participant. The researcher discussed the document and the overall idea of the experiment with the participant, who was invited to ask questions. The participant was informed that they could stop the experiment at any time for any reason and still receive full compensation ($20 cash) for participating in an experimental trial. To avoid biasing the experiment, participants were not told the specific research hypotheses.
6.3.3 Prescreening Survey Participants were administered a prescreening survey on their demographics (i.e., age, gender, ethnicity, driving experience, highest level of education, and prior experience with driving simulators) and questions in the following areas: •
Vision – Good vision was crucial for this experiment. Participants were asked if they used corrective glasses or contact lenses while driving. Their abilities to see the driving environment clearly and to read visual instructions (displayed on the screen) to stop driving were confirmed.
•
Simulator sickness – Participants with previous driving simulation experience were asked about any simulator sickness that they experienced. If they had previously experienced simulator sickness, they were encouraged not to participate in the experiment.
•
Motion sickness – Participants were surveyed about any kind of motion sickness they had experienced in the past. If an individual had a strong tendency towards any kind of motion sickness, they were encouraged not to participate in the experiment.
60
6.3.4 Calibration Drive After completing the prescreening survey, participants performed a 5-minute calibration drive. The overall purpose of this drive was to acclimate participants to the mechanics of the vehicle and the virtual reality of the simulator and to determine if they were prone to simulator sickness. Once seated in the vehicle for the test drive, participants were allowed to adjust the seat, rearview mirror, and steering wheel to maximize comfort and performance while driving. They were instructed to drive and follow all traffic laws as they normally would. According to Zhao et al. (2015), effective calibration drives introduce the participant to two primary roadway characteristics in the simulator environment: acceleration and deceleration on a stretch of roadway, and turning at intersections. Figure 6.13 shows the standard calibration drive that was developed for this experiment, which included the two elements. The environment of the calibration drive was simple and did not include any roadside development. Large yellow billboards with arrows were used to instruct the driver on which way to turn at intersections. Before the calibration drive, participants were instructed to follow the arrows on the billboards.
Figure 6.13: Layout of calibration drive Figure 6.14 shows views from two locations in the calibration drive. Participants who reported experiencing simulator sickness during or after the calibration drive were not allowed to continue in the experimental drives.
61
Figure 6.14: Screenshots of calibration drive in the simulation. Left: Approach to a curve near the beginning of the drive. Right: Approach to a signalized intersection, where a yellow billboard indicates a right turn
6.3.5 Eye-Tracking Calibration After the calibration drive was completed, researchers equipped participants with a headmounted eye tracker. Participants were directed to look at different locations on a calibration image projected on the forward screen of the driving simulator (Figure 6.15). If the eye-tracking equipment was unable to perform the calibration, which depended on eye position and other physical attributes of the participant, then the experiment was not continued.
Figure 6.15: Eye-tracking calibration image
62
6.3.6 Experimental Drive After the motorist’s eyes were calibrated to the driving simulator screens, they were given brief instructions about the test environment and the tasks that they were required to perform. The experiment was divided into six tracks. At the completion of each experimental drive, the researcher instructed the participant to stop the vehicle and ascertained whether the participant was experiencing simulator sickness. The virtual driving course (six tracks) was designed to take 30 to 40 minutes to complete (total).
6.3.7 Postdrive Survey As the final step of the experiment, drivers were asked to respond to several questions in a postdrive digital survey, which included questions about their previous responses to the indications, and how they interpreted each indication. The survey asked if the participant had experienced any simulator sickness during the experiment. The entire experiment, including the consent process, eye-tracker calibration, and postdrive questionnaire, lasted about 1 hour.
6.4 DATA REDUCTION Eye-tracking and simulator data for participants were carefully reduced to extract fixations on the various AOIs. The following sections describe these processing procedures.
6.4.1 Eye-Tracking Data After collecting participants’ eye-movement data, fixation and dwell data were analyzed by AOI polygons with the ETAnalysis software suite. For this process, researchers watched each video segment that included a right turn at an intersection (24 per participant). These video segments were cropped to the length of time that the driver entered the turning bay (generally 6–50 seconds). Next, researchers drew AOI polygons on individual video frames in a sequence separated by intervals of approximately 5–10 frames. Once the researcher manually situated each AOI, an “anchor” was created within the software. Distance and size differences of the AOIs between these anchors were interpolated by the ETAnalysis software to ensure that all fixations and dwells on the AOIs (i.e., signal display, pedestrians) were captured. Researchers analyzed motorist’s eye-tracking data starting from the point when the participant entered the turning bay at the intersection and continued until the participant completed the right-turn maneuver. Therefore, all of the objects of concern related to the current research questions appeared before the right-turning maneuvers were completed. Figure 6.16 is a screenshot of the ETAnalysis software, presenting an example of a video that has been coded with AOIs. At this particular moment, the motorist was fixating on the traffic signal. This figure also includes heat maps (orange-yellow circular patterns) of the pedestrian traffic signal head AOI with a walk interval in the motorist’s field of view.
63
Figure 6.16: Example of a participant fixation pattern on the signal head (two AOIs) Figure 6.17 presents an example of a participant fixating on a pedestrian in the conflicting crosswalk, with the AOI shown at the center of the figure by the blue rectangle and red crosshairs. This figure exemplifies a complex driving scenario in which the motorist, before turning right at the intersection, had to scan for all three key AOIs examined in this study (traffic signal display, pedestrian signal head, and crossing pedestrian in the conflicting crosswalk).
Figure 6.17: Example of a participant fixation pattern on pedestrian (three AOIs) 64
Once the AOIs were coded for each individual video file, output spreadsheets of all fixations and dwells for each AOI were produced by using the ETAnalysis software. Fixations and dwells outside the coded AOIs were universally defined as OUTSIDE and were not analyzed further. Researchers exported these data files and imported them into different analysis packages (e.g., Microsoft Excel and RStudio) for further analysis.
6.4.2 Simulator Data Simulator data were collected from the driving simulator and SimObserver platform during the experiment. A complete data file was generated for each participant for each of the six experimental drives. Files, including collected video data and all output of vehicle characteristics (e.g., lateral position and velocity), were opened in the Data Distillery (version 1.34) software suite, which provided quantitative outputs (numerical and graphical) in combination with the recorded video. Figure 6.18 shows the SimObserver video output in conjunction with numerical data (right side) and graphical representations of data in columns (bottom). In the Data Distillery program, the 24 right-turning maneuvers for each participant were located using the video data. Velocity and acceleration/deceleration data from the simulator corresponding to the maneuver were segmented from the point where the participant entered the turning bay until the right-turning maneuver was reached.
Figure 6.18: Screenshot of Data Distillery software interface (identifiable participant information was removed)
6.5
RESULTS AND ANALYSIS
This section presents the results of the simulator experiment. The first subsection describes the participant demographics. The next subsection describes the collection of driver decision-making 65
data from the SimObserver platform. Finally, the results from the analysis of visual attention and driver decision making are described. 6.5.1
Participants
Study participants were recruited from the community in and around Corvallis, Oregon. 6.5.1.1
Summary Statistics
The simulator study included 52 participants (27 men), ~10% (3 women, 2 men) of whom reported simulator sickness and did not complete the experiment (Table 6.3). All responses from participants who exhibited simulator sickness were excluded from the analyzed data set. Failure to calibrate the experimental equipment accurately resulted in the loss of data for one additional participant. The final analyzed sample population comprised 46 participants (M age = 30.9, SD age = 11.7) who completed the experiment and had complete simulator data including 21 women (M age = 29.3, SD age = 11.8) and 25 men (M age = 32.3, SD age = 11.7), and 43 of whom also had complete eye-tracking data. Table 6.3: Summary of Participant Population POPULATION TOTAL 52 (100%) Total enrolled 5 (10%) Simulator sickness (%) Experimental calibration 1 (2%) issue (%) 46 (88%) Final analyzed sample (%) 6.5.1.2
MEN 27 (52%) 2 (40%)
WOMEN 25 (48%) 3 (60%)
0 (0%)
1 (100%)
25 (54%)
21 (46%)
Demographics
Every effort was made to recruit a representative sample of the Oregon driving public. Table 6.4 summarizes the demographic information that was recorded during the prescreening survey for the 46 participants with complete information. All participants were licensed drivers who resided in the state of Oregon (not necessarily Oregon licensed).
66
Table 6.4: Participant Demographics POSSIBLE RESPONSES 1–5 years 6–10 years How many years have you 11–15 years been licensed? 16–20 years More than 20 years 1 time per week How often do 2–4 times per week you drive in a 5–10 times per week week? More than 10 times per week 0–5,000 miles 5,000–10,000 miles How many 10,000–15,000 miles miles did you 15,000–20,000 miles drive last year? More than 20,000 miles Glasses What corrective Contacts lenses do you wear while None driving? Yes Do you experience motion No sickness?
NUMBER OF PARTICIPANTS 18 12 3 6 7 6 9 12
PERCENTAGE OF PARTICIPANTS 39% 26% 7% 13% 15% 13% 20% 26%
19
41%
15 12 9 8
33% 26% 20% 17%
2
4%
0 7
0% 15%
39
85%
6
13%
40
87%
6.5.2 Postdrive Survey Results After participants completed the driving simulator experiment, they were asked to complete a short postdrive survey. Participants were asked what the appropriate response would be if they were turning right and saw either the steady red arrow or flashing yellow arrow indication. Three choices were offered: they could turn right cautiously without stopping; they could turn right after coming to a complete stop and finding a gap; or they could stop and wait for the green before turning. Driver responses are summarized in Figure 6.19. Overall, 65% of participants demonstrated poor comprehension of the correct action when faced with a steady red arrow display, whereas 78% of participants chose the correct response to the flashing yellow arrow display.
67
In Oregon, if you are turning right and you see the steady red arrow or the flashing yellow arrow displays, what would be the appropriate response for both?
Choices
Turn Right Cautiously
Turn Right, But First Stop FYA RA Stop and Wait for Green
0%
20%
40%
60%
80%
100%
Percentage of Responses
Figure 6.19: Participant comprehensions of the flashing yellow arrow (FYA) and red arrow (RA) indications When participants were asked to indicate how confident they were in their responses, 76% of participants were very confident that they must stop and wait until they received a green indication before turning when presented a steady red right-turn arrow. However, their confidence level was confident to very confident that they could turn right cautiously without stopping on the flashing yellow arrow display while turning right. One of the independent variables explored in this experiment was whether participants saw any differences between the circular red and red arrow indications and between the circular green and flashing yellow arrow indications. Participants were asked in the postdrive survey to identify whether or not they perceived these indications as the same while turning right in Oregon. Responses to these questions (Table 6.5) showed that 61% of participant thought the circular red and the steady red arrow indications meant different things, whereas 39% of participants thought that they had the same meaning. Most participants (80%) indicated that the circular green and flashing yellow arrow indications had different meanings.
68
Table 6.5: Postdrive Survey Question Responses QUESTION OPTIONS As a driver in Oregon, do the Yes steady circular red ball and the steady red arrow for the rightturning maneuver mean the same No to you? As a driver in Oregon, do the Yes flashing yellow arrow and the steady circular green ball for the right-turning maneuver mean the No same to you? 6.5.3
RESPONSES
PERCENTAGE
18
39%
28
61%
9
20%
37
80%
Driver Decision Making
Video data provide information on participant behavior that can facilitate interpretation of quantitative data. Decisions in response to each signal display were extracted from the SimOberver data. Data were carefully reviewed and classified into three categories: Correct, Partially correct, and Incorrect/Missing, based on established criteria shown in Table 6.6. The much of the coding convention for the actions observed in the simulator experiment followed the coding criteria of survey (Table 4.4).
69
Table 6.6: Error Coding of Narrative for the Simulator Experiments DISPLAY INDICATION CORRECT PARTIALLY CORRECT Steady Circular Green
Turn right with caution after yielding to pedestrians (if present) in the crosswalk
Steady Green Arrow
Turn right without stopping, recognizing that the SGA indicates a protected rightturn movement
Steady Circular Red and Steady Red Arrow
Come to a complete stop (vehicle speed < 1 mph) and complete the turn when they find a safe gap Turn right with caution after yielding to pedestrians (if present) in crosswalk
Flashing Yellow Arrow
INCORRECT
Turn right without checking for pedestrians even though the walk indication was displayed (or) not checking before turning but stopping once they saw a pedestrian Check for pedestrians and turn right (or) slow down and check for pedestrians and other cross traffic but did not recognize the protected movement in either case Turn right without coming to a complete stop (vehicle speed > 1 mph)
Stop before turning (vehicle speed < 1 mph) to check for pedestrians (or) A crash with pedestrian
Turn right without caution (vehicle speed > 15 mph) (or) not yielding when necessary
Stop before turning (vehicle speed < 1 mph) to check for pedestrians (or) remain stopped until the green indication
Stop before turning (vehicle speed < 1 mph) to check for pedestrians
Stop and remain stopped until the green indication
For the steady circular green (SCG), to be coded as correct, participants must turn right after yielding to pedestrians (if present) in the crosswalk. Partially correct actions resulted from drivers turning right without checking for pedestrians even though the walk indication was displayed, or not checking before turning but stopping once they saw a pedestrian. Incorrect 70
actions resulted from either drivers coming to a complete stop (vehicle speed < 1 mph) to check for pedestrians, or a crash with a pedestrian. Driver responses were coded as correct if they turned right without stopping, recognizing that the steady green arrow (SGA) indicates a protected right-turn movement. A response was coded partially incorrect if drivers slowed down or checked for pedestrians and other conflicting movements before turning. The incorrect action resulted from drivers coming to a complete stop (vehicle speed < 1 mph) while checking for pedestrians. The coding convention followed for the steady circular red (SCR) and the steady red arrow (SRS) indications was the same, as according to Oregon law the expected correct action for both display indications is the same for right turns. Responses were classified as correct if participants came to a complete stop (vehicle speed < 1 mph) and completed the turn when a safe gap was selected. Partially correct actions resulted from drivers making the right turn without coming to a complete stop. Responses were coded as incorrect if participants came to a complete stop and waited for a green indication. For the flashing yellow arrow (FYA), driver responses were coded as correct if they exhibited caution while turning and yielded to pedestrians when present, stopping when necessary. Partially correct actions resulted from drivers not turning right with caution (vehicle speed > 15 mph) or not yielding when necessary. Responses were coded as incorrect if drivers came to a complete stop (vehicle speed < 1 mph) before turning, or if they remained stopped until the signal display became green. For example, one participant stopped in response to the red arrow, rejected an acceptable gap, and waited for the green arrow indication before performing a right turn (Figure 6.20). This participant incorrectly interpreted the red arrow indication. After completing the experimental drives, the participant remarked to the researcher that they thought the right-turn red arrow display was as restrictive as it is for left turns.
Figure 6.20: Screenshot of Simobserver platform
71
6.5.3.1
Steady Circular Green
Responses were coded following the convention outlined in Table 6.6. Nine crashes or near-miss crashes were observed during 96 right turns in the 50- and 100-ft right-turn bays in the presence of pedestrians. Table 6.7 summarizes results for participants’ comprehension of the steady circular green indication. The circular green indication was evaluated in the simulator in scenarios with turning bay lengths of 50 and 100 ft, with or without a pedestrian present. Proportions of correct actions ranged from 67% (50-ft bay with pedestrian) to 76% (100-ft bay without pedestrian), partially correct actions ranged from 11% (100-ft bay without pedestrian) to 33% (50-ft bay with pedestrian), and incorrect actions ranged from 0% (50- or 100-ft bay with pedestrian) to 13% (100-ft bay without pedestrian). In general, the steady circular green indication was well understood, and all incorrect driver actions resulted in fail-safe driver responses. Table 6.7: Steady Circular Green Circular Responses RESPONSE 50 - NO PED 50 - PED 100 - NO PED 6.5% 6.5% 13.0% Incorrect 19.6% 26.1% 10.9% Partially Correct 73.9% 67.4% 76.1% Correct 6.5.3.2
100 - PED 13.0% 13.1% 73.9%
Steady Green Arrow
Responses were coded following the convention outlined in Table 6.6. Table 6.8 summarizes participants’ responses to the steady green arrow indication. Table 6.8: Steady Green Arrow Responses RESPONSE 50 - NO PED 0.0% Incorrect 2.2% Partially Correct 97.8% Correct
50 - PED 2.2% 13.0% 84.8%
100 - NO PED 0.0% 15.2% 84.8%
High proportions (85%-98%) of correct actions were observed among simulator participants responding to the steady green arrow indication in the 50-ft turning bay with and without a pedestrian and in the 100-ft turning bay without a pedestrian. Due to a programming error, the results from the 100-ft turning bay scenario with pedestrians are not reported. Responses deemed as partially correct varied between 2% and 15%. Low levels of incorrect responses were observed (0%-2%). These results indicate that the steady green arrow indication was well understood by driving simulator participants.
72
6.5.3.3
Steady Circular Red
Responses were coded following the convention outlined in Table 6.6. Table 6.9 summarizes participants’ responses to the steady circular red indication. Table 6.9: Steady Circular Red Responses RESPONSE 50 - NO PED 50 - PED 30.4% 28.3% Incorrect 21.7% Partially Correct 6.5% 63.0% 50.0% Correct
100 - NO PED 26.1% 21.7% 52.2%
100 - PED 30.4% 6.5% 63.0%
Proportions of correct actions to the steady circular red indication ranged from 50% (50ft bay without pedestrian) to 63% (50-ft bay without pedestrian, 100-ft bay with pedestrian), partially correct actions ranged from 7% to 22%, and incorrect actions ranged from 26% to 30%. Proportions of incorrect actions were higher for this indication than for the steady circular green indication. These errors resulted in fail-safe responses; although they may result in efficiency losses, they are not expected to reduce safety. 6.5.3.4
Steady Red Arrow
Responses were coded following the convention outlined in Table 6.6. Table 6.10 summarizes responses to this indication. Table 6.10: Steady Red Arrow Responses RESPONSE 50 - NO PED 50 - PED 69.6% 67.4% Incorrect 2.2% Partially Correct 6.5% 23.9% 30.4% Correct
100 - NO PED 65.2% 8.7% 26.1%
100 - PED 67.1% 0.0% 32.6%
Drivers exhibited low levels of correct responses (24%-33%) to the steady red arrow indication due to their tendency to stop and wait for a steady green indication before proceeding. These actions are considered fail-safe responses and they affect efficiency more directly than safety. The proportions between the different tested scenarios were similar. 6.5.3.5
Flashing Yellow Arrow with Pedestrian Walk Interval (FYAW)
Responses were coded following the convention outlined in Table 6.6. Table 6.11 summarizes responses to the flashing yellow arrow during the pedestrian walk interval. Table 6.11: Responses to Flashing Yellow Arrow during Pedestrian Walk Interval RESPONSE 50 - NO PED 50 - PED 100 - NO PED 100 - PED 15.2% 0.0% 10.9% 0.0% Incorrect 2.2% 0.0% 2.2% Partially Correct 0.0% 84.8% 97.8% 89.1% 97.8% Correct 73
High levels of correct responses (85% to 98%) were observed with the flashing yellow arrow indication. Incorrect actions were primarily observed with the 50- and 100-ft turning bay scenarios with no pedestrian present (11% and 15%). The high proportion of correct responses indicates that the flashing yellow arrow indication is well understood by Oregon drivers. 6.5.3.6
Flashing Yellow Arrow with Pedestrian Clearance Interval (FYAC)
The same coding as was used with FYAW was used to identify drivers’ responses to the flashing yellow arrow during the pedestrian clearance interval. High levels of comprehension of the FWAC were observed (89%-98%). Incorrect actions were observed for the 50- and 100-ft turning bay scenarios when no pedestrian was present. Thus, using a flashing yellow arrow during the pedestrian clearance interval resulted in drivers correctly responding in most situations. Table 6.12 summarizes participants’ comprehension of this indication. Table 6.12: Responses to Flashing Yellow Arrow during Pedestrian Clearance Interval RESPONSE 50 - NO PED 50 - PED 100 - NO PED 100 - PED 6.5% 0.0% 4.4% 0.0% Incorrect 0.0% 2.2% 0.0% 10.9% Partially Correct 93.5% 97.8% 95.6% 89.1% Correct 6.5.4
Visual Attention
Visual attention data were gathered and reduced from the ASL Mobile Eye XG for the 43 participants with complete eye-tracking data. This section organizes eye-tracking results by the three independent variables: signal indication type and active display, right-turn bay length, and presence of pedestrians. Data were analyzed in R Studio version 1.0.153 and SAS 9.4. 6.5.4.1
Average Total Fixation Duration (ATFD)
The total fixation duration (TFD) within a prescribed AOI was used to measure motorists’ visual attention on different targets. For each right-turning maneuver, the number and duration (in seconds) of participants’ fixations on AOIs were recorded, with a TFD of 0 seconds indicating that the participant did not look at the target. The ATFD was calculated by averaging all participants’ total fixations on an AOI. 6.5.4.2
Descriptive Data and Statistical Analysis
The type of signal display, turning bay length, and presence of a pedestrian in the conflicting crosswalk may influence motorists’ visual attention while turning right. Therefore, these factors were included as independent variables. Although other factors (e.g., motorists’ experience level, age, pedestrian conspicuity) may also influence motorists’ visual search task at an intersection, those factors are outside the scope of this study. The first independent variable, “type of signal indication,” had six levels: 1) circular green, 2) circular red, 3) green arrow, 4) red arrow, 5) flashing yellow arrow with walk interval on (FYAW), and 6) flashing yellow arrow with clearance interval on (FYAC). The second independent variable, “turning bay length,” had two levels: 1) 50 ft 74
and 2) 100 ft. The last independent variable, “presence of a conflicting pedestrian in the crosswalk,” had two levels: 1) no pedestrian and 2) one pedestrian walking towards the participant in the conflicting crosswalk. One of the primary dependent variables of this experiment was the visual attention of motorists during the right-turn maneuver. Fixation data for different AOIs were statistically analyzed by using SAS (version 9.4). Each participant completed 24 right-turning maneuvers and received all possible combinations of independent variables (6×2×2). Therefore, repeated measure analysis of variance (ANOVA) was used to statistically determine if there were any differences in the variances between levels with respect to display type, AOI, or bay length. Pairwise comparisons of estimated marginal means TFDs were calculated with Tukey’s Honest Significant Difference (HSD) test. In the following subsections, the TFD is used as a performance metric to compare the visual distribution potential between levels of signal indications. 6.5.4.3
Circular Red and Red Arrow Indications
Participants’ responses while turning right at intersections operating the MUTCD standard circular red signal configuration were compared to responses at intersections with the steady red arrow display. The TFD was calculated for each participant during each right-turning maneuver. Right turns were sorted by signal indication type, turning bay length, and AOI (signal and pedestrian signal). Data were visualized as boxplots of TFD disaggregated by three independent variables. Figure 6.21 shows TFD values for AOIs at intersections with no pedestrians, 50- and 100-ft turning bay lengths, and circular red and red arrow indications. Median TFDs ranged from 0.10 to 3.91 seconds. The highest median TFD occurred with the red arrow indication in the 50-ft bay while fixating on the signal head. The lowest median TFD occurred with the circular red indication in the 50-ft bay while fixating on the pedestrian signal head.
Figure 6.21: Boxplots of TFD disaggregated by signal type 75
Table 6.13 contains descriptive statistics for TFDs for each independent variable. Repeated-measures ANOVA, used to determine whether mean TFD was affected by the three factors (signal type, bay length, and AOI), revealed a significant effect of the listed variables on the mean TFD (F (7, 343) = 29.40, p < 0.001). Table 6.13: Descriptive Statistics for TFD by Signal Type, Bay Length, and AOI AOI SIGNAL PEDSIG SIGNAL TYPE CR RA CR RA BAY LENGTH 100ft 50ft 100ft 50ft 100ft 50ft 100ft 50ft 1.26 3.60 3.91 0.23 0.10 0.43 0.47 MEDIAN 1.51 1.60 3.85 4.98 0.44 0.28 0.92 0.81 DESCRIPTIVE MEAN 1.93 STATISTICS 1.78 1.49 3.24 3.82 0.67 0.37 1.28 1.10 STD DEV
Tukey’s HSD post hoc pairwise comparison was performed to understand differences between group means. Results at the 95% confidence level (Table 6.14) showed a significant difference between mean TFDs for interactions with the circular red, 50-ft bay, and signal AOI (M = 1.60, SD = 1.49) compared to interactions with the red arrow, 50-ft bay, and signal AOI (M = 4.98, SD = 3.82). A significant difference between mean TFDs was found for interactions with the circular red, 100-ft bay, and signal AOI (M = 1.93, SD = 1.78) compared to interactions with the red arrow, 100-ft bay, and signal AOI (M = 3.85, SD = 3.24). No significant difference between mean TFDs was observed when the AOI was changed for the same combination of the other two factors. Table 6.14: Tukey’s HSD Multiple Comparisons by Signal Type, Bay Length, and AOI 95% CI MEAN P COMPARISON DIFFERENCE LOWER UPPER CR:50:Signal vs. 3.37
